Edit model card

DistilBERTFINAL_ctxSentence_TRAIN_essays_TEST_NULL_second_train_set_null_False

This model is a fine-tuned version of cardiffnlp/twitter-roberta-base on the None dataset. It achieves the following results on the evaluation set:

  • Loss: 0.7680
  • Precision: 0.9838
  • Recall: 0.6632
  • F1: 0.7923
  • Accuracy: 0.6624

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: 1e-05
  • train_batch_size: 32
  • eval_batch_size: 32
  • 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
No log 1.0 130 0.2980 0.9315 0.9533 0.9423 0.9081
No log 2.0 260 0.2053 0.9537 0.9626 0.9581 0.9338
No log 3.0 390 0.1873 0.9464 0.9907 0.9680 0.9485
0.3064 4.0 520 0.1811 0.9585 0.9720 0.9652 0.9449
0.3064 5.0 650 0.1887 0.9587 0.9766 0.9676 0.9485

Framework versions

  • Transformers 4.15.0
  • Pytorch 1.10.1+cu113
  • Datasets 1.18.0
  • Tokenizers 0.10.3
Downloads last month
6