Hugging Face's logo

liaad
/
NER_harem_albertina-100m-portuguese-ptpt-encoder

Token Classification
Transformers
Safetensors
deberta
generated_from_trainer
Eval Results
Inference Endpoints
Model card Files Community
NER_harem_albertina-100m-portuguese-ptpt-encoder / README.md
arubenruben's picture
arubenruben
End of training
3b16337 verified
preview code
|
raw history blame contribute delete
No virus
3.91 kB
metadata
license: mit
base_model: PORTULAN/albertina-100m-portuguese-ptpt-encoder
tags:
  - generated_from_trainer
datasets:
  - harem
metrics:
  - precision
  - recall
  - f1
  - accuracy
model-index:
  - name: NER_harem_albertina-100m-portuguese-ptpt-encoder
    results:
      - task:
          name: Token Classification
          type: token-classification
        dataset:
          name: harem
          type: harem
          config: default
          split: test
          args: default
        metrics:
          - name: Precision
            type: precision
            value: 0.67216673903604
          - name: Recall
            type: recall
            value: 0.725398313027179
          - name: F1
            type: f1
            value: 0.6977687626774848
          - name: Accuracy
            type: accuracy
            value: 0.9532056132627089

NER_harem_albertina-100m-portuguese-ptpt-encoder

This model is a fine-tuned version of PORTULAN/albertina-100m-portuguese-ptpt-encoder on the harem dataset. It achieves the following results on the evaluation set:

  • Loss: 0.2583
  • Precision: 0.6722
  • Recall: 0.7254
  • F1: 0.6978
  • Accuracy: 0.9532

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: 3e-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: 300

Training results

Training Loss Epoch Step Validation Loss Precision Recall F1 Accuracy
No log 1.0 16 0.5322 0.0212 0.0117 0.0151 0.8615
No log 2.0 32 0.3238 0.4230 0.4981 0.4575 0.9110
No log 3.0 48 0.2460 0.5006 0.6007 0.5461 0.9369
No log 4.0 64 0.2240 0.5526 0.6396 0.5930 0.9414
No log 5.0 80 0.2088 0.5498 0.6340 0.5889 0.9492
No log 6.0 96 0.2068 0.5884 0.6645 0.6241 0.9496
No log 7.0 112 0.2253 0.5906 0.6720 0.6287 0.9481
No log 8.0 128 0.2115 0.6245 0.6874 0.6545 0.9516
No log 9.0 144 0.2187 0.6546 0.7062 0.6794 0.9533
No log 10.0 160 0.2398 0.6432 0.7020 0.6713 0.9495
No log 11.0 176 0.2554 0.6653 0.7043 0.6843 0.9526
No log 12.0 192 0.2397 0.6777 0.7212 0.6988 0.9529
No log 13.0 208 0.2565 0.6778 0.7207 0.6986 0.9531
No log 14.0 224 0.2700 0.6586 0.7142 0.6853 0.9506
No log 15.0 240 0.2700 0.7009 0.7259 0.7132 0.9544
No log 16.0 256 0.2688 0.6761 0.7240 0.6993 0.9532
No log 17.0 272 0.2741 0.7132 0.7343 0.7236 0.9558
No log 18.0 288 0.2732 0.6740 0.7132 0.6931 0.9530
No log 19.0 304 0.2745 0.7094 0.7310 0.7201 0.9550
No log 20.0 320 0.2583 0.6722 0.7254 0.6978 0.9532

Framework versions

  • Transformers 4.38.2
  • Pytorch 2.2.1+cu121
  • Datasets 2.18.0
  • Tokenizers 0.15.2