metadata

license: mit
tags:
  - generated_from_trainer
datasets:
  - wikiann
metrics:
  - precision
  - recall
  - f1
  - accuracy
model-index:
  - name: deberta-finetuned-ner-connll-late-stop
    results:
      - task:
          name: Token Classification
          type: token-classification
        dataset:
          name: wikiann
          type: wikiann
          config: en
          split: train
          args: en
        metrics:
          - name: Precision
            type: precision
            value: 0.830192600803658
          - name: Recall
            type: recall
            value: 0.8470945850417079
          - name: F1
            type: f1
            value: 0.8385584324702589
          - name: Accuracy
            type: accuracy
            value: 0.9228861596598961

deberta-finetuned-ner-connll-late-stop

This model is a fine-tuned version of microsoft/deberta-base on the wikiann dataset. It achieves the following results on the evaluation set:

Loss: 0.5259
Precision: 0.8302
Recall: 0.8471
F1: 0.8386
Accuracy: 0.9229

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: 7

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
0.3408	1.0	1875	0.3639	0.7462	0.7887	0.7669	0.8966
0.2435	2.0	3750	0.2933	0.8104	0.8332	0.8217	0.9178
0.1822	3.0	5625	0.3034	0.8147	0.8388	0.8266	0.9221
0.1402	4.0	7500	0.3667	0.8275	0.8474	0.8374	0.9235
0.1013	5.0	9375	0.4290	0.8285	0.8448	0.8366	0.9227
0.0677	6.0	11250	0.4914	0.8259	0.8473	0.8365	0.9231
0.0439	7.0	13125	0.5259	0.8302	0.8471	0.8386	0.9229

Framework versions

Transformers 4.22.1
Pytorch 1.12.1+cu113
Datasets 2.4.0
Tokenizers 0.12.1