distilbert-base-uncased-finetuned-clinic-best

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

• Loss: 0.1552
• Accuracy: 0.9410

Model description

“The parameters are as follows:

{'num_train_epoch': 9, 'alpha': 0.16484386886358915, 'temperature': 2}

These were obtained using grid search with Optuna.

Training and evaluation data

Clinic dataset from 'datasets'

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 2e-05
• train_batch_size: 48
• eval_batch_size: 48
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 9

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
No log	1.0	318	0.8641	0.7094
1.1497	2.0	636	0.4002	0.8703
1.1497	3.0	954	0.2381	0.9113
0.3873	4.0	1272	0.1893	0.9313
0.2047	5.0	1590	0.1716	0.9339
0.2047	6.0	1908	0.1632	0.9387
0.165	7.0	2226	0.1589	0.9394
0.151	8.0	2544	0.1561	0.94
0.151	9.0	2862	0.1552	0.9410

Framework versions

• Transformers 4.47.0
• Pytorch 2.5.1+cu124
• Datasets 3.2.0
• Tokenizers 0.21.0