consejo-ner

This model is a fine-tuned version of dccuchile/distilbert-base-spanish-uncased on an unknown dataset. It achieves the following results on the evaluation set:

  • Loss: 0.3066
  • Precision: 0.7241
  • Recall: 0.6774
  • F1: 0.7
  • Accuracy: 0.9313

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: 16
  • eval_batch_size: 16
  • seed: 42
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • num_epochs: 20

Training results

Training Loss Epoch Step Validation Loss Precision Recall F1 Accuracy
No log 1.0 15 1.5724 0.0 0.0 0.0 0.6985
No log 2.0 30 1.3540 0.0 0.0 0.0 0.6985
No log 3.0 45 1.0972 0.0 0.0 0.0 0.7099
No log 4.0 60 0.8615 0.5833 0.2258 0.3256 0.7672
No log 5.0 75 0.7381 0.5 0.3548 0.4151 0.8244
No log 6.0 90 0.6111 0.5556 0.4839 0.5172 0.8473
No log 7.0 105 0.5353 0.5185 0.4516 0.4828 0.8550
No log 8.0 120 0.4786 0.5769 0.4839 0.5263 0.8626
No log 9.0 135 0.4493 0.5357 0.4839 0.5085 0.8817
No log 10.0 150 0.4269 0.4839 0.4839 0.4839 0.8779
No log 11.0 165 0.3977 0.5938 0.6129 0.6032 0.8931
No log 12.0 180 0.3669 0.5161 0.5161 0.5161 0.8969
No log 13.0 195 0.3437 0.6786 0.6129 0.6441 0.9237
No log 14.0 210 0.3389 0.6786 0.6129 0.6441 0.9198
No log 15.0 225 0.3249 0.6786 0.6129 0.6441 0.9198
No log 16.0 240 0.3102 0.6897 0.6452 0.6667 0.9275
No log 17.0 255 0.3094 0.6667 0.6452 0.6557 0.9275
No log 18.0 270 0.3159 0.7 0.6774 0.6885 0.9198
No log 19.0 285 0.3094 0.7241 0.6774 0.7 0.9313
No log 20.0 300 0.3066 0.7241 0.6774 0.7 0.9313

Framework versions

  • Transformers 4.26.1
  • Pytorch 1.13.1+cu116
  • Datasets 2.9.0
  • Tokenizers 0.13.2
Downloads last month
14
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.