<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->
address_parsing_test_1
This model is a fine-tuned version of csebuetnlp/banglabert on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.0207
- Precision: 0.9953
- Recall: 0.9957
- F1: 0.9955
- Accuracy: 0.9963
Model description
More information needed
Intended uses & limitations
More information needed
Training and evaluation data
More information needed
Training procedure
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 2e-05
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 5
Training results
| Training Loss | Epoch | Step | Validation Loss | Precision | Recall | F1 | Accuracy |
|---|---|---|---|---|---|---|---|
| 0.0669 | 1.0 | 1266 | 0.0168 | 0.9890 | 0.9913 | 0.9902 | 0.9937 |
| 0.02 | 2.0 | 2532 | 0.0132 | 0.9953 | 0.9961 | 0.9957 | 0.9967 |
| 0.0122 | 3.0 | 3798 | 0.0177 | 0.9949 | 0.9955 | 0.9952 | 0.9962 |
| 0.0068 | 4.0 | 5064 | 0.0181 | 0.9951 | 0.9957 | 0.9954 | 0.9963 |
| 0.0048 | 5.0 | 6330 | 0.0207 | 0.9953 | 0.9957 | 0.9955 | 0.9963 |
Framework versions
- Transformers 4.24.0
- Pytorch 1.12.1+cu113
- Datasets 2.7.1
- Tokenizers 0.13.2
