using custom data but gives run time error . Please suggest

[rajeev12111]

---

RuntimeError Traceback (most recent call last)
Cell In[18], line 15
10 config.verbose = False # If verbose == True, PyABSA will output the model strcture and seversal processed data examples
11 config.notice = (
12 "This is an training example for aspect term extraction" # for memos usage
13 )
---> 15 trainer = ATEPC.ATEPCTrainer(
16 config=config,
17 dataset=my_dataset,
18 from_checkpoint="english", # if you want to resume training from our pretrained checkpoints, you can pass the checkpoint name here
19 auto_device=DeviceTypeOption.AUTO, # use cuda if available
20 checkpoint_save_mode=ModelSaveOption.SAVE_MODEL_STATE_DICT, # save state dict only instead of the whole model
21 load_aug=False, # there are some augmentation dataset for integrated datasets, you use them by setting load_aug=True to improve performance
22 )

File /opt/conda/lib/python3.10/site-packages/pyabsa/tasks/AspectTermExtraction/trainer/atepc_trainer.py:69, in ATEPCTrainer.init(self, config, dataset, from_checkpoint, checkpoint_save_mode, auto_device, path_to_save, load_aug)
64 self.config.task_code = TaskCodeOption.Aspect_Term_Extraction_and_Classification
65 self.config.task_name = TaskNameOption().get(
66 TaskCodeOption.Aspect_Term_Extraction_and_Classification
67 )
---> 69 self._run()

File /opt/conda/lib/python3.10/site-packages/pyabsa/framework/trainer_class/trainer_template.py:241, in Trainer._run(self)
239 self.config.seed = s
240 if self.config.checkpoint_save_mode:
--> 241 model_path.append(self.training_instructor(self.config).run())
242 else:
243 # always return the last trained model if you don't save trained model
244 model = self.inference_model_class(
245 checkpoint=self.training_instructor(self.config).run()
246 )

File /opt/conda/lib/python3.10/site-packages/pyabsa/tasks/AspectTermExtraction/instructor/atepc_instructor.py:794, in ATEPCTrainingInstructor.run(self)
793 def run(self):
--> 794 return self._train(criterion=None)

File /opt/conda/lib/python3.10/site-packages/pyabsa/framework/instructor_class/instructor_template.py:357, in BaseTrainingInstructor._train(self, criterion)
354 pass
356 # Resume training from a previously trained model
--> 357 self._resume_from_checkpoint()
359 # Initialize the learning rate scheduler if warmup_step is specified
360 if self.config.warmup_step >= 0:

File /opt/conda/lib/python3.10/site-packages/pyabsa/framework/instructor_class/instructor_template.py:455, in BaseTrainingInstructor._resume_from_checkpoint(self)
451 self.model.module.load_state_dict(
452 torch.load(state_dict_path[0])
453 )
454 else:
--> 455 self.model.load_state_dict(
456 torch.load(
457 state_dict_path[0], map_location=self.config.device
458 )
459 )
460 self.model.config = self.config
461 self.model.to(self.config.device)

File /opt/conda/lib/python3.10/site-packages/torch/nn/modules/module.py:2041, in Module.load_state_dict(self, state_dict, strict)
2036 error_msgs.insert(
2037 0, 'Missing key(s) in state_dict: {}. '.format(
2038 ', '.join('"{}"'.format(k) for k in missing_keys)))
2040 if len(error_msgs) > 0:
-> 2041 raise RuntimeError('Error(s) in loading state_dict for {}:\n\t{}'.format(
2042 self.class.name, "\n\t".join(error_msgs)))
2043 return _IncompatibleKeys(missing_keys, unexpected_keys)

RuntimeError: Error(s) in loading state_dict for FAST_LCF_ATEPC:
Unexpected key(s) in state_dict: "bert4global.embeddings.position_ids".

[rajeev12111]

from pyabsa import DatasetItem
my_dataset = DatasetItem("/kaggle/working/integrated_datasets/atepc_datasets/custom.apc.train.txt.atepc")

from pyabsa import ModelSaveOption, DeviceTypeOption
import warnings

warnings.filterwarnings("ignore")

config.batch_size = 16
config.patience = 2
config.log_step = -1
config.seed = [1, 2, 3]
config.verbose = False # If verbose == True, PyABSA will output the model strcture and seversal processed data examples
config.notice = (
"This is an training example for aspect term extraction" # for memos usage
)

trainer = ATEPC.ATEPCTrainer(
config=config,
dataset=my_dataset,
from_checkpoint="english", # if you want to resume training from our pretrained checkpoints, you can pass the checkpoint name here
auto_device=DeviceTypeOption.AUTO, # use cuda if available
checkpoint_save_mode=ModelSaveOption.SAVE_MODEL_STATE_DICT, # save state dict only instead of the whole model
load_aug=False, # there are some augmentation dataset for integrated datasets, you use them by setting load_aug=True to improve performance
)

suggest to correct and debug the code

[yangheng95]

try pip install pyabsa -U for V2 and pip install pyabsa<2.0 -Ufor V1

[rajeev12111]

[2023-10-08 12:23:53] (2.3.4) Set Model Device: cuda:0
[2023-10-08 12:23:53] (2.3.4) Device Name: Tesla T4
Downloading (…)lve/main/config.json: 100%
579/579 [00:00<00:00, 45.9kB/s]
2023-10-08 12:23:53,894 INFO: PyABSA version: 2.3.4
2023-10-08 12:23:53,897 INFO: Transformers version: 4.33.0
2023-10-08 12:23:53,898 INFO: Torch version: 2.0.0+cuda11.8
2023-10-08 12:23:53,899 INFO: Device: Tesla T4
2023-10-08 12:23:53,900 INFO: custom.apc.train.txt.atepc in the trainer is not a exact path, will search dataset in current working directory
2023-10-08 12:23:53,901 INFO: Warning! auto_evaluate=True, however cannot find test set using for evaluating!
Downloading (…)okenizer_config.json: 100%
52.0/52.0 [00:00<00:00, 3.95kB/s]
Downloading spm.model: 100%
2.46M/2.46M [00:00<00:00, 43.8MB/s]
Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained.
Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained.
convert examples to features: 100%|██████████| 100/100 [00:00<00:00, 741.28it/s]
2023-10-08 12:23:56,319 INFO: Dataset Label Details: {'NEGATIVE': 26, 'POSITIVE': 63, 'NEUTRAL': 11, 'Sum': 100}

Downloading pytorch_model.bin: 100%
371M/371M [00:01<00:00, 309MB/s]
2023-10-08 12:24:00,702 INFO: Save cache dataset to fast_lcf_atepc.custom.apc.train.txt.atepc.dataset.99144ac14eaf5168ddd19ec0b56a2cbcfccfe564762e5fb2e81dac19326b747d.cache
2023-10-08 12:24:07,532 INFO: cuda memory allocated:764963840
2023-10-08 12:24:07,533 INFO: ABSADatasetsVersion:None --> Calling Count:0
2023-10-08 12:24:07,535 INFO: IOB_label_to_index:{'B-ASP': 1, 'I-ASP': 2, 'O': 3, '[CLS]': 4, '[SEP]': 5} --> Calling Count:1
2023-10-08 12:24:07,536 INFO: MV:<metric_visualizer.metric_visualizer.MetricVisualizer object at 0x7f89f822bb80> --> Calling Count:0
2023-10-08 12:24:07,537 INFO: PyABSAVersion:2.3.4 --> Calling Count:1
2023-10-08 12:24:07,538 INFO: SRD:3 --> Calling Count:200
2023-10-08 12:24:07,539 INFO: TorchVersion:2.0.0+cuda11.8 --> Calling Count:1
2023-10-08 12:24:07,540 INFO: TransformersVersion:4.33.0 --> Calling Count:1
2023-10-08 12:24:07,541 INFO: auto_device:True --> Calling Count:2
2023-10-08 12:24:07,542 INFO: batch_size:16 --> Calling Count:3
2023-10-08 12:24:07,543 INFO: cache_dataset:True --> Calling Count:1
2023-10-08 12:24:07,544 INFO: checkpoint_save_mode:1 --> Calling Count:4
2023-10-08 12:24:07,545 INFO: cross_validate_fold:-1 --> Calling Count:0
2023-10-08 12:24:07,546 INFO: dataset_file:{'train': ['/kaggle/working/integrated_datasets/atepc_datasets/custom.apc.train.txt.atepc'], 'test': [], 'valid': []} --> Calling Count:5
2023-10-08 12:24:07,547 INFO: dataset_name:custom.apc.train.txt.atepc --> Calling Count:3
2023-10-08 12:24:07,548 INFO: device:cuda:0 --> Calling Count:4
2023-10-08 12:24:07,548 INFO: device_name:Tesla T4 --> Calling Count:1
2023-10-08 12:24:07,549 INFO: dropout:0.5 --> Calling Count:1
2023-10-08 12:24:07,550 INFO: dynamic_truncate:True --> Calling Count:200
2023-10-08 12:24:07,551 INFO: embed_dim:768 --> Calling Count:0
2023-10-08 12:24:07,551 INFO: evaluate_begin:0 --> Calling Count:0
2023-10-08 12:24:07,552 INFO: from_checkpoint:english --> Calling Count:0
2023-10-08 12:24:07,553 INFO: gradient_accumulation_steps:1 --> Calling Count:3
2023-10-08 12:24:07,554 INFO: hidden_dim:768 --> Calling Count:6
2023-10-08 12:24:07,554 INFO: index_to_IOB_label:{1: 'B-ASP', 2: 'I-ASP', 3: 'O', 4: '[CLS]', 5: '[SEP]'} --> Calling Count:0
2023-10-08 12:24:07,555 INFO: index_to_label:{0: 'NEGATIVE', 1: 'NEUTRAL', 2: 'POSITIVE'} --> Calling Count:1
2023-10-08 12:24:07,556 INFO: inference_model:None --> Calling Count:0
2023-10-08 12:24:07,557 INFO: initializer:xavier_uniform_ --> Calling Count:0
2023-10-08 12:24:07,557 INFO: l2reg:1e-05 --> Calling Count:2
2023-10-08 12:24:07,558 INFO: label_list:['B-ASP', 'I-ASP', 'O', '[CLS]', '[SEP]'] --> Calling Count:1
2023-10-08 12:24:07,559 INFO: label_to_index:{'NEGATIVE': 0, 'NEUTRAL': 1, 'POSITIVE': 2} --> Calling Count:0
2023-10-08 12:24:07,559 INFO: lcf:cdw --> Calling Count:0
2023-10-08 12:24:07,560 INFO: learning_rate:2e-05 --> Calling Count:1
2023-10-08 12:24:07,561 INFO: load_aug:False --> Calling Count:1
2023-10-08 12:24:07,562 INFO: log_step:-1 --> Calling Count:0
2023-10-08 12:24:07,563 INFO: logger:<Logger fast_lcf_atepc (INFO)> --> Calling Count:8
2023-10-08 12:24:07,563 INFO: max_seq_len:80 --> Calling Count:701
2023-10-08 12:24:07,564 INFO: model:<class 'pyabsa.tasks.AspectTermExtraction.models.lcf.fast_lcf_atepc.FAST_LCF_ATEPC'> --> Calling Count:5
2023-10-08 12:24:07,565 INFO: model_name:fast_lcf_atepc --> Calling Count:102
2023-10-08 12:24:07,566 INFO: model_path_to_save:checkpoints --> Calling Count:3
2023-10-08 12:24:07,566 INFO: notice:This is an training example for aspect term extraction --> Calling Count:0
2023-10-08 12:24:07,567 INFO: num_epoch:10 --> Calling Count:1
2023-10-08 12:24:07,568 INFO: num_labels:6 --> Calling Count:2
2023-10-08 12:24:07,569 INFO: optimizer:adamw --> Calling Count:2
2023-10-08 12:24:07,569 INFO: output_dim:3 --> Calling Count:1
2023-10-08 12:24:07,570 INFO: overwrite_cache:False --> Calling Count:0
2023-10-08 12:24:07,571 INFO: path_to_save:None --> Calling Count:1
2023-10-08 12:24:07,572 INFO: patience:2 --> Calling Count:0
2023-10-08 12:24:07,573 INFO: pretrained_bert:microsoft/deberta-v3-base --> Calling Count:5
2023-10-08 12:24:07,573 INFO: save_mode:1 --> Calling Count:0
2023-10-08 12:24:07,574 INFO: seed:1 --> Calling Count:6
2023-10-08 12:24:07,575 INFO: sep_indices:2 --> Calling Count:0
2023-10-08 12:24:07,576 INFO: show_metric:False --> Calling Count:0
2023-10-08 12:24:07,576 INFO: spacy_model:en_core_web_sm --> Calling Count:1
2023-10-08 12:24:07,577 INFO: srd_alignment:True --> Calling Count:0
2023-10-08 12:24:07,578 INFO: task_code:ATEPC --> Calling Count:1
2023-10-08 12:24:07,579 INFO: task_name:Aspect Term Extraction and Polarity Classification --> Calling Count:0
2023-10-08 12:24:07,579 INFO: use_amp:False --> Calling Count:1
2023-10-08 12:24:07,580 INFO: use_bert_spc:True --> Calling Count:0
2023-10-08 12:24:07,581 INFO: use_syntax_based_SRD:False --> Calling Count:100
2023-10-08 12:24:07,582 INFO: verbose:False --> Calling Count:0
2023-10-08 12:24:07,582 INFO: warmup_step:-1 --> Calling Count:0
2023-10-08 12:24:07,583 INFO: window:lr --> Calling Count:0
2023-10-08 12:24:07,587 INFO: cuda memory allocated:764963840
2023-10-08 12:24:07,588 INFO: ABSADatasetsVersion:None --> Calling Count:0
2023-10-08 12:24:07,589 INFO: IOB_label_to_index:{'B-ASP': 1, 'I-ASP': 2, 'O': 3, '[CLS]': 4, '[SEP]': 5} --> Calling Count:1
2023-10-08 12:24:07,590 INFO: MV:<metric_visualizer.metric_visualizer.MetricVisualizer object at 0x7f89f822bb80> --> Calling Count:0
2023-10-08 12:24:07,591 INFO: PyABSAVersion:2.3.4 --> Calling Count:1
2023-10-08 12:24:07,591 INFO: SRD:3 --> Calling Count:200
2023-10-08 12:24:07,592 INFO: TorchVersion:2.0.0+cuda11.8 --> Calling Count:1
2023-10-08 12:24:07,593 INFO: TransformersVersion:4.33.0 --> Calling Count:1
2023-10-08 12:24:07,594 INFO: auto_device:True --> Calling Count:3
2023-10-08 12:24:07,594 INFO: batch_size:16 --> Calling Count:3
2023-10-08 12:24:07,595 INFO: cache_dataset:True --> Calling Count:1
2023-10-08 12:24:07,596 INFO: checkpoint_save_mode:1 --> Calling Count:4
2023-10-08 12:24:07,597 INFO: cross_validate_fold:-1 --> Calling Count:1
2023-10-08 12:24:07,597 INFO: dataset_file:{'train': ['/kaggle/working/integrated_datasets/atepc_datasets/custom.apc.train.txt.atepc'], 'test': [], 'valid': []} --> Calling Count:5
2023-10-08 12:24:07,598 INFO: dataset_name:custom.apc.train.txt.atepc --> Calling Count:3
2023-10-08 12:24:07,599 INFO: device:cuda:0 --> Calling Count:8
2023-10-08 12:24:07,600 INFO: device_name:Tesla T4 --> Calling Count:1
2023-10-08 12:24:07,600 INFO: dropout:0.5 --> Calling Count:1
2023-10-08 12:24:07,601 INFO: dynamic_truncate:True --> Calling Count:200
2023-10-08 12:24:07,602 INFO: embed_dim:768 --> Calling Count:0
2023-10-08 12:24:07,603 INFO: evaluate_begin:0 --> Calling Count:0
2023-10-08 12:24:07,603 INFO: from_checkpoint:english --> Calling Count:0
2023-10-08 12:24:07,604 INFO: gradient_accumulation_steps:1 --> Calling Count:3
2023-10-08 12:24:07,605 INFO: hidden_dim:768 --> Calling Count:6
2023-10-08 12:24:07,606 INFO: index_to_IOB_label:{1: 'B-ASP', 2: 'I-ASP', 3: 'O', 4: '[CLS]', 5: '[SEP]'} --> Calling Count:0
2023-10-08 12:24:07,606 INFO: index_to_label:{0: 'NEGATIVE', 1: 'NEUTRAL', 2: 'POSITIVE'} --> Calling Count:1
2023-10-08 12:24:07,607 INFO: inference_model:None --> Calling Count:0
2023-10-08 12:24:07,608 INFO: initializer:xavier_uniform --> Calling Count:0
2023-10-08 12:24:07,609 INFO: l2reg:1e-05 --> Calling Count:2
2023-10-08 12:24:07,609 INFO: label_list:['B-ASP', 'I-ASP', 'O', '[CLS]', '[SEP]'] --> Calling Count:1
2023-10-08 12:24:07,610 INFO: label_to_index:{'NEGATIVE': 0, 'NEUTRAL': 1, 'POSITIVE': 2} --> Calling Count:0
2023-10-08 12:24:07,611 INFO: lcf:cdw --> Calling Count:0
2023-10-08 12:24:07,612 INFO: learning_rate:2e-05 --> Calling Count:1
2023-10-08 12:24:07,612 INFO: load_aug:False --> Calling Count:1
2023-10-08 12:24:07,613 INFO: log_step:-1 --> Calling Count:0
2023-10-08 12:24:07,614 INFO: logger:<_Logger fast_lcf_atepc (INFO)> --> Calling Count:8
2023-10-08 12:24:07,615 INFO: max_seq_len:80 --> Calling Count:701
2023-10-08 12:24:07,615 INFO: model:<class 'pyabsa.tasks.AspectTermExtraction.models.lcf.fast_lcf_atepc.FAST_LCF_ATEPC'> --> Calling Count:5
2023-10-08 12:24:07,616 INFO: model_name:fast_lcf_atepc --> Calling Count:102
2023-10-08 12:24:07,617 INFO: model_path_to_save:checkpoints --> Calling Count:3
2023-10-08 12:24:07,618 INFO: notice:This is an training example for aspect term extraction --> Calling Count:0
2023-10-08 12:24:07,619 INFO: num_epoch:10 --> Calling Count:1
2023-10-08 12:24:07,620 INFO: num_labels:6 --> Calling Count:2
2023-10-08 12:24:07,621 INFO: optimizer:adamw --> Calling Count:2
2023-10-08 12:24:07,621 INFO: output_dim:3 --> Calling Count:1
2023-10-08 12:24:07,622 INFO: overwrite_cache:False --> Calling Count:0
2023-10-08 12:24:07,623 INFO: path_to_save:None --> Calling Count:1
2023-10-08 12:24:07,624 INFO: patience:2 --> Calling Count:0
2023-10-08 12:24:07,624 INFO: pretrained_bert:microsoft/deberta-v3-base --> Calling Count:5
2023-10-08 12:24:07,625 INFO: save_mode:1 --> Calling Count:0
2023-10-08 12:24:07,626 INFO: seed:1 --> Calling Count:6
2023-10-08 12:24:07,627 INFO: sep_indices:2 --> Calling Count:0
2023-10-08 12:24:07,627 INFO: show_metric:False --> Calling Count:0
2023-10-08 12:24:07,628 INFO: spacy_model:en_core_web_sm --> Calling Count:1
2023-10-08 12:24:07,629 INFO: srd_alignment:True --> Calling Count:0
2023-10-08 12:24:07,630 INFO: task_code:ATEPC --> Calling Count:1
2023-10-08 12:24:07,630 INFO: task_name:Aspect Term Extraction and Polarity Classification --> Calling Count:0
2023-10-08 12:24:07,631 INFO: tokenizer:DebertaV2TokenizerFast(name_or_path='microsoft/deberta-v3-base', vocab_size=128000, model_max_length=1000000000000000019884624838656, is_fast=True, padding_side='right', truncation_side='right', special_tokens={'bos_token': '[CLS]', 'eos_token': '[SEP]', 'unk_token': '[UNK]', 'sep_token': '[SEP]', 'pad_token': '[PAD]', 'cls_token': '[CLS]', 'mask_token': '[MASK]'}, clean_up_tokenization_spaces=True) --> Calling Count:0
2023-10-08 12:24:07,632 INFO: use_amp:False --> Calling Count:1
2023-10-08 12:24:07,633 INFO: use_bert_spc:True --> Calling Count:0
2023-10-08 12:24:07,633 INFO: use_syntax_based_SRD:False --> Calling Count:100
2023-10-08 12:24:07,634 INFO: verbose:False --> Calling Count:1
2023-10-08 12:24:07,635 INFO: warmup_step:-1 --> Calling Count:0
2023-10-08 12:24:07,636 INFO: window:lr --> Calling Count:0
[2023-10-08 12:24:07] (2.3.4) Downloading checkpoint:english
[2023-10-08 12:24:07] (2.3.4) Notice: The pretrained model are used for testing, it is recommended to train the model on your own custom datasets
Downloading checkpoint: 579MB [00:01, 333.35MB/s]
Find zipped checkpoint: ./checkpoints/ATEPC_ENGLISH_CHECKPOINT/fast_lcf_atepc_English_cdw_apcacc_82.36_apcf1_81.89_atef1_75.43.zip, unzipping

Done.
[2023-10-08 12:24:16] (2.3.4) If the auto-downloading failed, please download it via browser: https://huggingface.co/spaces/yangheng/PyABSA/resolve/main/checkpoints/English/ATEPC/fast_lcf_atepc_English_cdw_apcacc_82.36_apcf1_81.89_atef1_75.43.zip
2023-10-08 12:24:16,518 INFO: Checkpoint downloaded at: checkpoints/ATEPC_ENGLISH_CHECKPOINT/fast_lcf_atepc_English_cdw_apcacc_82.36_apcf1_81.89_atef1_75.43

RuntimeError Traceback (most recent call last)
Cell In[8], line 15
10 config.verbose = False # If verbose == True, PyABSA will output the model strcture and seversal processed data examples
11 config.notice = (
12 "This is an training example for aspect term extraction" # for memos usage
13 )
---> 15 trainer = ATEPC.ATEPCTrainer(
16 config=config,
17 dataset=my_dataset,
18 from_checkpoint="english", # if you want to resume training from our pretrained checkpoints, you can pass the checkpoint name here
19 auto_device=DeviceTypeOption.AUTO, # use cuda if available
20 checkpoint_save_mode=ModelSaveOption.SAVE_MODEL_STATE_DICT, # save state dict only instead of the whole model
21 load_aug=False, # there are some augmentation dataset for integrated datasets, you use them by setting load_aug=True to improve performance
22 )

File /opt/conda/lib/python3.10/site-packages/pyabsa/tasks/AspectTermExtraction/trainer/atepc_trainer.py:69, in ATEPCTrainer.init(self, config, dataset, from_checkpoint, checkpoint_save_mode, auto_device, path_to_save, load_aug)
64 self.config.task_code = TaskCodeOption.Aspect_Term_Extraction_and_Classification
65 self.config.task_name = TaskNameOption().get(
66 TaskCodeOption.Aspect_Term_Extraction_and_Classification
67 )
---> 69 self._run()

File /opt/conda/lib/python3.10/site-packages/pyabsa/framework/trainer_class/trainer_template.py:241, in Trainer._run(self)
239 self.config.seed = s
240 if self.config.checkpoint_save_mode:
--> 241 model_path.append(self.training_instructor(self.config).run())
242 else:
243 # always return the last trained model if you don't save trained model
244 model = self.inference_model_class(
245 checkpoint=self.training_instructor(self.config).run()
246 )

File /opt/conda/lib/python3.10/site-packages/pyabsa/tasks/AspectTermExtraction/instructor/atepc_instructor.py:794, in ATEPCTrainingInstructor.run(self)
793 def run(self):
--> 794 return self._train(criterion=None)

File /opt/conda/lib/python3.10/site-packages/pyabsa/framework/instructor_class/instructor_template.py:357, in BaseTrainingInstructor._train(self, criterion)
354 pass
356 # Resume training from a previously trained model
--> 357 self._resume_from_checkpoint()
359 # Initialize the learning rate scheduler if warmup_step is specified
360 if self.config.warmup_step >= 0:

File /opt/conda/lib/python3.10/site-packages/pyabsa/framework/instructor_class/instructor_template.py:455, in BaseTrainingInstructor._resume_from_checkpoint(self)
451 self.model.module.load_state_dict(
452 torch.load(state_dict_path[0])
453 )
454 else:
--> 455 self.model.load_state_dict(
456 torch.load(
457 state_dict_path[0], map_location=self.config.device
458 )
459 )
460 self.model.config = self.config
461 self.model.to(self.config.device)

File /opt/conda/lib/python3.10/site-packages/torch/nn/modules/module.py:2041, in Module.load_state_dict(self, state_dict, strict)
2036 error_msgs.insert(
2037 0, 'Missing key(s) in state_dict: {}. '.format(
2038 ', '.join('"{}"'.format(k) for k in missing_keys)))
2040 if len(error_msgs) > 0:
-> 2041 raise RuntimeError('Error(s) in loading state_dict for {}:\n\t{}'.format(
2042 self.class.name, "\n\t".join(error_msgs)))
2043 return _IncompatibleKeys(missing_keys, unexpected_keys)

RuntimeError: Error(s) in loading state_dict for FAST_LCF_ATEPC:
Unexpected key(s) in state_dict: "bert4global.embeddings.position_ids". I have installed pip install pyabsa -U for V2 but showing this type of error. . I stuck now. Thank you

[yangheng95]

try pip install pyabsa==2.3.4rc0

[rajeev12111]

Thank you so much for quick response . I am not able to Fine-Tuning Training for Your Own Model . I stuck now. I am not able to debug the error from this pip install pyabsa==2.3.4rc0 . Please suggest the script for Fine-Tuning Training for Your Own Model.

Thank you

[rajeev12111]

Please suggest to debugging

[yangheng95]

Thank you so much for quick response . I am not able to Fine-Tuning Training for Your Own Model . I stuck now. I am not able to debug the error from this pip install pyabsa==2.3.4rc0 . Please suggest the script for Fine-Tuning Training for Your Own Model.

Thank you

I cannot well understand your reply, does the error seem like the previuous?

[rajeev12111]

try pip install pyabsa==2.3.4rc0

Thank you so much for quick response . I am not able to Fine-Tuning Training for Your Own Model . I stuck now. I am not able to debug the error from this pip install pyabsa==2.3.4rc0 . Please suggest the script for Fine-Tuning Training for Your Own Model.
Thank you

I cannot well understand your reply, does the error seem like the previuous?

Yes, the error seem like the previous.

[yangheng95]

Sorry I cannot reproduce the error. Did you restart the Kernal after the update?

[rajeev12111]

my code script here
from pyabsa import DatasetItem
my_dataset = DatasetItem("/kaggle/working/integrated_datasets/atepc_datasets/custom.apc.train.txt.atepc")

from pyabsa import ModelSaveOption, DeviceTypeOption
import warnings

warnings.filterwarnings("ignore")

config.batch_size = 16
config.patience = 2
config.log_step = -1
config.seed = [1, 2, 3]
config.verbose = False # If verbose == True, PyABSA will output the model strcture and seversal processed data examples
config.notice = (
"This is an training example for aspect term extraction" # for memos usage
)

trainer = ATEPC.ATEPCTrainer(
config=config,
dataset=my_dataset,
from_checkpoint="english", # if you want to resume training from our pretrained checkpoints, you can pass the checkpoint name here
auto_device=DeviceTypeOption.AUTO, # use cuda if available
checkpoint_save_mode=ModelSaveOption.SAVE_MODEL_STATE_DICT, # save state dict only instead of the whole model
load_aug=False, # there are some augmentation dataset for integrated datasets, you use them by setting load_aug=True to improve performance
)

[2023-10-09 09:22:34] (2.3.4rc0) Set Model Device: cuda:0
[2023-10-09 09:22:34] (2.3.4rc0) Device Name: Tesla T4
Downloading (…)lve/main/config.json: 100%
579/579 [00:00<00:00, 16.9kB/s]
2023-10-09 09:22:35,125 INFO: PyABSA version: 2.3.4rc0
2023-10-09 09:22:35,130 INFO: Transformers version: 4.33.0
2023-10-09 09:22:35,131 INFO: Torch version: 2.0.0+cuda11.8
2023-10-09 09:22:35,132 INFO: Device: Tesla T4
2023-10-09 09:22:35,133 INFO: custom.apc.train.txt.atepc in the trainer is not a exact path, will search dataset in current working directory
2023-10-09 09:22:35,134 INFO: Warning! auto_evaluate=True, however cannot find test set using for evaluating!
Downloading (…)okenizer_config.json: 100%
52.0/52.0 [00:00<00:00, 1.39kB/s]
Downloading spm.model: 100%
2.46M/2.46M [00:00<00:00, 24.5MB/s]
Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained.
Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained.
convert examples to features: 100%|██████████| 100/100 [00:00<00:00, 763.36it/s]
2023-10-09 09:22:37,566 INFO: Dataset Label Details: {'NEUTRAL': 11, 'POSITIVE': 63, 'NEGATIVE': 26, 'Sum': 100}

Downloading pytorch_model.bin: 100%
371M/371M [00:01<00:00, 218MB/s]
2023-10-09 09:22:41,963 INFO: Save cache dataset to fast_lcf_atepc.custom.apc.train.txt.atepc.dataset.675d8f23577290a43bfabc0f4f088e6703da03520d63e199a6202e1497c62f1b.cache
2023-10-09 09:22:47,111 INFO: cuda memory allocated:764963840
2023-10-09 09:22:47,113 INFO: ABSADatasetsVersion:None --> Calling Count:0
2023-10-09 09:22:47,114 INFO: IOB_label_to_index:{'B-ASP': 1, 'I-ASP': 2, 'O': 3, '[CLS]': 4, '[SEP]': 5} --> Calling Count:1
2023-10-09 09:22:47,115 INFO: MV:<metric_visualizer.metric_visualizer.MetricVisualizer object at 0x7ee80f108cd0> --> Calling Count:0
2023-10-09 09:22:47,116 INFO: PyABSAVersion:2.3.4rc0 --> Calling Count:1
2023-10-09 09:22:47,118 INFO: SRD:3 --> Calling Count:200
2023-10-09 09:22:47,119 INFO: TorchVersion:2.0.0+cuda11.8 --> Calling Count:1
2023-10-09 09:22:47,120 INFO: TransformersVersion:4.33.0 --> Calling Count:1
2023-10-09 09:22:47,121 INFO: auto_device:True --> Calling Count:2
2023-10-09 09:22:47,122 INFO: batch_size:16 --> Calling Count:3
2023-10-09 09:22:47,123 INFO: cache_dataset:True --> Calling Count:1
2023-10-09 09:22:47,124 INFO: checkpoint_save_mode:1 --> Calling Count:4
2023-10-09 09:22:47,125 INFO: cross_validate_fold:-1 --> Calling Count:0
2023-10-09 09:22:47,126 INFO: dataset_file:{'train': ['/kaggle/working/integrated_datasets/atepc_datasets/custom.apc.train.txt.atepc'], 'test': [], 'valid': []} --> Calling Count:5
2023-10-09 09:22:47,127 INFO: dataset_name:custom.apc.train.txt.atepc --> Calling Count:3
2023-10-09 09:22:47,128 INFO: device:cuda:0 --> Calling Count:4
2023-10-09 09:22:47,128 INFO: device_name:Tesla T4 --> Calling Count:1
2023-10-09 09:22:47,129 INFO: dropout:0.5 --> Calling Count:1
2023-10-09 09:22:47,130 INFO: dynamic_truncate:True --> Calling Count:200
2023-10-09 09:22:47,131 INFO: embed_dim:768 --> Calling Count:0
2023-10-09 09:22:47,132 INFO: evaluate_begin:0 --> Calling Count:0
2023-10-09 09:22:47,133 INFO: from_checkpoint:english --> Calling Count:0
2023-10-09 09:22:47,134 INFO: gradient_accumulation_steps:1 --> Calling Count:3
2023-10-09 09:22:47,135 INFO: hidden_dim:768 --> Calling Count:6
2023-10-09 09:22:47,136 INFO: index_to_IOB_label:{1: 'B-ASP', 2: 'I-ASP', 3: 'O', 4: '[CLS]', 5: '[SEP]'} --> Calling Count:0
2023-10-09 09:22:47,137 INFO: index_to_label:{0: 'NEGATIVE', 1: 'NEUTRAL', 2: 'POSITIVE'} --> Calling Count:1
2023-10-09 09:22:47,137 INFO: inference_model:None --> Calling Count:0
2023-10-09 09:22:47,138 INFO: initializer:xavier_uniform_ --> Calling Count:0
2023-10-09 09:22:47,139 INFO: l2reg:1e-05 --> Calling Count:2
2023-10-09 09:22:47,140 INFO: label_list:['B-ASP', 'I-ASP', 'O', '[CLS]', '[SEP]'] --> Calling Count:1
2023-10-09 09:22:47,141 INFO: label_to_index:{'NEGATIVE': 0, 'NEUTRAL': 1, 'POSITIVE': 2} --> Calling Count:0
2023-10-09 09:22:47,142 INFO: lcf:cdw --> Calling Count:0
2023-10-09 09:22:47,143 INFO: learning_rate:2e-05 --> Calling Count:1
2023-10-09 09:22:47,144 INFO: load_aug:False --> Calling Count:1
2023-10-09 09:22:47,144 INFO: log_step:-1 --> Calling Count:0
2023-10-09 09:22:47,145 INFO: logger:<Logger fast_lcf_atepc (INFO)> --> Calling Count:8
2023-10-09 09:22:47,146 INFO: max_seq_len:80 --> Calling Count:701
2023-10-09 09:22:47,147 INFO: model:<class 'pyabsa.tasks.AspectTermExtraction.models.lcf.fast_lcf_atepc.FAST_LCF_ATEPC'> --> Calling Count:5
2023-10-09 09:22:47,148 INFO: model_name:fast_lcf_atepc --> Calling Count:102
2023-10-09 09:22:47,149 INFO: model_path_to_save:checkpoints --> Calling Count:3
2023-10-09 09:22:47,149 INFO: notice:This is an training example for aspect term extraction --> Calling Count:0
2023-10-09 09:22:47,150 INFO: num_epoch:10 --> Calling Count:1
2023-10-09 09:22:47,151 INFO: num_labels:6 --> Calling Count:2
2023-10-09 09:22:47,152 INFO: optimizer:adamw --> Calling Count:2
2023-10-09 09:22:47,153 INFO: output_dim:3 --> Calling Count:1
2023-10-09 09:22:47,154 INFO: overwrite_cache:False --> Calling Count:0
2023-10-09 09:22:47,155 INFO: path_to_save:None --> Calling Count:1
2023-10-09 09:22:47,156 INFO: patience:2 --> Calling Count:0
2023-10-09 09:22:47,157 INFO: pretrained_bert:microsoft/deberta-v3-base --> Calling Count:5
2023-10-09 09:22:47,158 INFO: save_mode:1 --> Calling Count:0
2023-10-09 09:22:47,159 INFO: seed:1 --> Calling Count:6
2023-10-09 09:22:47,160 INFO: sep_indices:2 --> Calling Count:0
2023-10-09 09:22:47,161 INFO: show_metric:False --> Calling Count:0
2023-10-09 09:22:47,161 INFO: spacy_model:en_core_web_sm --> Calling Count:1
2023-10-09 09:22:47,162 INFO: srd_alignment:True --> Calling Count:0
2023-10-09 09:22:47,163 INFO: task_code:ATEPC --> Calling Count:1
2023-10-09 09:22:47,164 INFO: task_name:Aspect Term Extraction and Polarity Classification --> Calling Count:0
2023-10-09 09:22:47,165 INFO: use_amp:False --> Calling Count:1
2023-10-09 09:22:47,166 INFO: use_bert_spc:True --> Calling Count:0
2023-10-09 09:22:47,167 INFO: use_syntax_based_SRD:False --> Calling Count:100
2023-10-09 09:22:47,168 INFO: verbose:False --> Calling Count:0
2023-10-09 09:22:47,169 INFO: warmup_step:-1 --> Calling Count:0
2023-10-09 09:22:47,170 INFO: window:lr --> Calling Count:0
2023-10-09 09:22:47,176 INFO: cuda memory allocated:764963840
2023-10-09 09:22:47,177 INFO: ABSADatasetsVersion:None --> Calling Count:0
2023-10-09 09:22:47,178 INFO: IOB_label_to_index:{'B-ASP': 1, 'I-ASP': 2, 'O': 3, '[CLS]': 4, '[SEP]': 5} --> Calling Count:1
2023-10-09 09:22:47,181 INFO: MV:<metric_visualizer.metric_visualizer.MetricVisualizer object at 0x7ee80f108cd0> --> Calling Count:0
2023-10-09 09:22:47,182 INFO: PyABSAVersion:2.3.4rc0 --> Calling Count:1
2023-10-09 09:22:47,185 INFO: SRD:3 --> Calling Count:200
2023-10-09 09:22:47,186 INFO: TorchVersion:2.0.0+cuda11.8 --> Calling Count:1
2023-10-09 09:22:47,187 INFO: TransformersVersion:4.33.0 --> Calling Count:1
2023-10-09 09:22:47,188 INFO: auto_device:True --> Calling Count:3
2023-10-09 09:22:47,189 INFO: batch_size:16 --> Calling Count:3
2023-10-09 09:22:47,192 INFO: cache_dataset:True --> Calling Count:1
2023-10-09 09:22:47,192 INFO: checkpoint_save_mode:1 --> Calling Count:4
2023-10-09 09:22:47,195 INFO: cross_validate_fold:-1 --> Calling Count:1
2023-10-09 09:22:47,196 INFO: dataset_file:{'train': ['/kaggle/working/integrated_datasets/atepc_datasets/custom.apc.train.txt.atepc'], 'test': [], 'valid': []} --> Calling Count:5
2023-10-09 09:22:47,197 INFO: dataset_name:custom.apc.train.txt.atepc --> Calling Count:3
2023-10-09 09:22:47,200 INFO: device:cuda:0 --> Calling Count:8
2023-10-09 09:22:47,201 INFO: device_name:Tesla T4 --> Calling Count:1
2023-10-09 09:22:47,202 INFO: dropout:0.5 --> Calling Count:1
2023-10-09 09:22:47,204 INFO: dynamic_truncate:True --> Calling Count:200
2023-10-09 09:22:47,205 INFO: embed_dim:768 --> Calling Count:0
2023-10-09 09:22:47,208 INFO: evaluate_begin:0 --> Calling Count:0
2023-10-09 09:22:47,209 INFO: from_checkpoint:english --> Calling Count:0
2023-10-09 09:22:47,210 INFO: gradient_accumulation_steps:1 --> Calling Count:3
2023-10-09 09:22:47,211 INFO: hidden_dim:768 --> Calling Count:6
2023-10-09 09:22:47,214 INFO: index_to_IOB_label:{1: 'B-ASP', 2: 'I-ASP', 3: 'O', 4: '[CLS]', 5: '[SEP]'} --> Calling Count:0
2023-10-09 09:22:47,215 INFO: index_to_label:{0: 'NEGATIVE', 1: 'NEUTRAL', 2: 'POSITIVE'} --> Calling Count:1
2023-10-09 09:22:47,216 INFO: inference_model:None --> Calling Count:0
2023-10-09 09:22:47,218 INFO: initializer:xavier_uniform --> Calling Count:0
2023-10-09 09:22:47,222 INFO: l2reg:1e-05 --> Calling Count:2
2023-10-09 09:22:47,223 INFO: label_list:['B-ASP', 'I-ASP', 'O', '[CLS]', '[SEP]'] --> Calling Count:1
2023-10-09 09:22:47,224 INFO: label_to_index:{'NEGATIVE': 0, 'NEUTRAL': 1, 'POSITIVE': 2} --> Calling Count:0
2023-10-09 09:22:47,225 INFO: lcf:cdw --> Calling Count:0
2023-10-09 09:22:47,226 INFO: learning_rate:2e-05 --> Calling Count:1
2023-10-09 09:22:47,228 INFO: load_aug:False --> Calling Count:1
2023-10-09 09:22:47,229 INFO: log_step:-1 --> Calling Count:0
2023-10-09 09:22:47,232 INFO: logger:<_Logger fast_lcf_atepc (INFO)> --> Calling Count:8
2023-10-09 09:22:47,233 INFO: max_seq_len:80 --> Calling Count:701
2023-10-09 09:22:47,234 INFO: model:<class 'pyabsa.tasks.AspectTermExtraction.models.lcf.fast_lcf_atepc.FAST_LCF_ATEPC'> --> Calling Count:5
2023-10-09 09:22:47,236 INFO: model_name:fast_lcf_atepc --> Calling Count:102
2023-10-09 09:22:47,237 INFO: model_path_to_save:checkpoints --> Calling Count:3
2023-10-09 09:22:47,240 INFO: notice:This is an training example for aspect term extraction --> Calling Count:0
2023-10-09 09:22:47,241 INFO: num_epoch:10 --> Calling Count:1
2023-10-09 09:22:47,242 INFO: num_labels:6 --> Calling Count:2
2023-10-09 09:22:47,243 INFO: optimizer:adamw --> Calling Count:2
2023-10-09 09:22:47,244 INFO: output_dim:3 --> Calling Count:1
2023-10-09 09:22:47,246 INFO: overwrite_cache:False --> Calling Count:0
2023-10-09 09:22:47,247 INFO: path_to_save:None --> Calling Count:1
2023-10-09 09:22:47,250 INFO: patience:2 --> Calling Count:0
2023-10-09 09:22:47,251 INFO: pretrained_bert:microsoft/deberta-v3-base --> Calling Count:5
2023-10-09 09:22:47,252 INFO: save_mode:1 --> Calling Count:0
2023-10-09 09:22:47,253 INFO: seed:1 --> Calling Count:6
2023-10-09 09:22:47,255 INFO: sep_indices:2 --> Calling Count:0
2023-10-09 09:22:47,256 INFO: show_metric:False --> Calling Count:0
2023-10-09 09:22:47,257 INFO: spacy_model:en_core_web_sm --> Calling Count:1
2023-10-09 09:22:47,259 INFO: srd_alignment:True --> Calling Count:0
2023-10-09 09:22:47,260 INFO: task_code:ATEPC --> Calling Count:1
2023-10-09 09:22:47,261 INFO: task_name:Aspect Term Extraction and Polarity Classification --> Calling Count:0
2023-10-09 09:22:47,264 INFO: tokenizer:DebertaV2TokenizerFast(name_or_path='microsoft/deberta-v3-base', vocab_size=128000, model_max_length=1000000000000000019884624838656, is_fast=True, padding_side='right', truncation_side='right', special_tokens={'bos_token': '[CLS]', 'eos_token': '[SEP]', 'unk_token': '[UNK]', 'sep_token': '[SEP]', 'pad_token': '[PAD]', 'cls_token': '[CLS]', 'mask_token': '[MASK]'}, clean_up_tokenization_spaces=True) --> Calling Count:0
2023-10-09 09:22:47,265 INFO: use_amp:False --> Calling Count:1
2023-10-09 09:22:47,266 INFO: use_bert_spc:True --> Calling Count:0
2023-10-09 09:22:47,268 INFO: use_syntax_based_SRD:False --> Calling Count:100
2023-10-09 09:22:47,269 INFO: verbose:False --> Calling Count:1
2023-10-09 09:22:47,272 INFO: warmup_step:-1 --> Calling Count:0
2023-10-09 09:22:47,273 INFO: window:lr --> Calling Count:0
[2023-10-09 09:22:47] (2.3.4rc0) Downloading checkpoint:english
[2023-10-09 09:22:47] (2.3.4rc0) Notice: The pretrained model are used for testing, it is recommended to train the model on your own custom datasets
Downloading checkpoint: 579MB [00:08, 68.25MB/s]
Find zipped checkpoint: ./checkpoints/ATEPC_ENGLISH_CHECKPOINT/fast_lcf_atepc_English_cdw_apcacc_82.36_apcf1_81.89_atef1_75.43.zip, unzipping

Done.
[2023-10-09 09:23:03] (2.3.4rc0) If the auto-downloading failed, please download it via browser: https://huggingface.co/spaces/yangheng/PyABSA/resolve/main/checkpoints/English/ATEPC/fast_lcf_atepc_English_cdw_apcacc_82.36_apcf1_81.89_atef1_75.43.zip
2023-10-09 09:23:03,138 INFO: Checkpoint downloaded at: checkpoints/ATEPC_ENGLISH_CHECKPOINT/fast_lcf_atepc_English_cdw_apcacc_82.36_apcf1_81.89_atef1_75.43
2023-10-09 09:23:03,828 INFO: Resume trainer from Checkpoint: checkpoints/ATEPC_ENGLISH_CHECKPOINT/fast_lcf_atepc_English_cdw_apcacc_82.36_apcf1_81.89_atef1_75.43!
2023-10-09 09:23:03,829 INFO: ***** Running training for Aspect Term Extraction and Polarity Classification *****
2023-10-09 09:23:03,831 INFO: Num examples = 100
2023-10-09 09:23:03,832 INFO: Batch size = 16
2023-10-09 09:23:03,834 INFO: Num steps = 60
Epoch: 0 | Smooth Loss: 1.0063: 100%|██████████| 7/7 [00:07<00:00, 1.08s/it]
Epoch: 1 | Smooth Loss: 0.8769: 100%|██████████| 7/7 [00:05<00:00, 1.28it/s]

TypeError Traceback (most recent call last)
Cell In[10], line 15
10 config.verbose = False # If verbose == True, PyABSA will output the model strcture and seversal processed data examples
11 config.notice = (
12 "This is an training example for aspect term extraction" # for memos usage
13 )
---> 15 trainer = ATEPC.ATEPCTrainer(
16 config=config,
17 dataset=my_dataset,
18 from_checkpoint="english", # if you want to resume training from our pretrained checkpoints, you can pass the checkpoint name here
19 auto_device=DeviceTypeOption.AUTO, # use cuda if available
20 checkpoint_save_mode=ModelSaveOption.SAVE_MODEL_STATE_DICT, # save state dict only instead of the whole model
21 load_aug=False, # there are some augmentation dataset for integrated datasets, you use them by setting load_aug=True to improve performance
22 )

File /opt/conda/lib/python3.10/site-packages/pyabsa/tasks/AspectTermExtraction/trainer/atepc_trainer.py:69, in ATEPCTrainer.init(self, config, dataset, from_checkpoint, checkpoint_save_mode, auto_device, path_to_save, load_aug)
64 self.config.task_code = TaskCodeOption.Aspect_Term_Extraction_and_Classification
65 self.config.task_name = TaskNameOption().get(
66 TaskCodeOption.Aspect_Term_Extraction_and_Classification
67 )
---> 69 self._run()

File /opt/conda/lib/python3.10/site-packages/pyabsa/framework/trainer_class/trainer_template.py:241, in Trainer._run(self)
239 self.config.seed = s
240 if self.config.checkpoint_save_mode:
--> 241 model_path.append(self.training_instructor(self.config).run())
242 else:
243 # always return the last trained model if you don't save trained model
244 model = self.inference_model_class(
245 checkpoint=self.training_instructor(self.config).run()
246 )

File /opt/conda/lib/python3.10/site-packages/pyabsa/tasks/AspectTermExtraction/instructor/atepc_instructor.py:794, in ATEPCTrainingInstructor.run(self)
793 def run(self):
--> 794 return self._train(criterion=None)

File /opt/conda/lib/python3.10/site-packages/pyabsa/framework/instructor_class/instructor_template.py:374, in BaseTrainingInstructor._train(self, criterion)
371 return self._k_fold_train_and_evaluate(criterion)
372 # Train and evaluate the model if there is only one validation dataloader
373 else:
--> 374 return self._train_and_evaluate(criterion)

File /opt/conda/lib/python3.10/site-packages/pyabsa/tasks/AspectTermExtraction/instructor/atepc_instructor.py:492, in ATEPCTrainingInstructor._train_and_evaluate(self, criterion)
489 if patience == 0:
490 break
--> 492 apc_result, ate_result = self._evaluate_acc_f1(self.test_dataloader)
494 if self.valid_set and self.test_set:
495 self.config.MV.log_metric(
496 self.config.model_name
497 + "-"
(...)
502 apc_result["apc_test_acc"],
503 )

File /opt/conda/lib/python3.10/site-packages/pyabsa/tasks/AspectTermExtraction/instructor/atepc_instructor.py:604, in ATEPCTrainingInstructor._evaluate_acc_f1(self, test_dataloader, eval_ATE, eval_APC)
601 self.model.eval()
602 label_map = {i: label for i, label in enumerate(self.config.label_list, 1)}
--> 604 for i_batch, batch in enumerate(test_dataloader):
605 (
606 input_ids_spc,
607 segment_ids,
(...)
614 lcf_cdw_vec,
615 ) = batch
617 input_ids_spc = input_ids_spc.to(self.config.device)

TypeError: 'NoneType' object is not iterable.

This is error, please debug it

[yangheng95]

You just forget to annotate the test dataset

[rajeev12111]

Thank you so much