About the inferring of aspect polarity about pyabsa HOT 7 CLOSED

I put all the generated state_dict into it, or each data set only needs to put the last one?

from pyabsa.

The script indentified the file name contained 'acc' as state-dict, see https://github.com/yangheng95/LC-ABSA/blob/c945a94e0f86116c5578245aa9ad36c46c7b9c4a/apc_infer.py#L149

And each file only need one.

from pyabsa.

Tell me if any problem continues

from pyabsa.

To match your example, I used dataset 'car'
like this
But this problem still exists

from pyabsa.

{
  "config1": {
    "model_name": "lcf_bert",
    "dataset": "car",
    "optimizer": "adam",
    "learning_rate": 0.00002,
    "dropout": 0.1,
    "l2reg": 0.00001,
    "num_epoch": 5,
    "batch_size": 4,
    "logdir": "logs",
    "pretrained_bert_name": "bert-base-chinese",
    "use_dual_bert": false,
    "use_bert_spc": true,
    "max_seq_len": 80,
    "SRD": 5,
    "sigma": 0.5,
    "lcf": "cdw",
    "lce": true,
    "lcp": true,
    "exp_rounds": 1
  }
}

from pyabsa.

Seems there is no seed in the file name. It is recommended to use the auto-generated state-dict file.
e.g. lce_bert_car_acc72.98_seed0seed

from pyabsa.

I am sorry to trouble you with such a simple question.
I can run it successfully, thank you!

from pyabsa.