dit-small_tobacco3482_kd_CEKD_t2.5_a0.5

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

• Loss: 3.8936
• Accuracy: 0.185
• Brier Loss: 0.8707
• Nll: 6.6284
• F1 Micro: 0.185
• F1 Macro: 0.0488
• Ece: 0.2527
• Aurc: 0.7434

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
• gradient_accumulation_steps: 16
• total_train_batch_size: 256
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 25

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Brier Loss	Nll	F1 Micro	F1 Macro	Ece	Aurc
No log	0.96	3	4.2363	0.06	0.9043	9.2962	0.06	0.0114	0.1758	0.9032
No log	1.96	6	4.1268	0.18	0.8887	6.8683	0.18	0.0305	0.2329	0.8055
No log	2.96	9	4.0044	0.18	0.8773	7.3055	0.18	0.0305	0.2510	0.8219
No log	3.96	12	3.9678	0.18	0.8851	7.2435	0.18	0.0305	0.2677	0.8214
No log	4.96	15	3.9645	0.185	0.8877	6.9806	0.185	0.0488	0.2757	0.7934
No log	5.96	18	3.9635	0.185	0.8853	6.9543	0.185	0.0488	0.2551	0.7812
No log	6.96	21	3.9564	0.185	0.8801	6.0556	0.185	0.0488	0.2515	0.7771
No log	7.96	24	3.9505	0.185	0.8772	6.0356	0.185	0.0488	0.2598	0.7724
No log	8.96	27	3.9435	0.185	0.8751	6.0288	0.185	0.0488	0.2590	0.7697
No log	9.96	30	3.9383	0.185	0.8742	6.0724	0.185	0.0488	0.2474	0.7712
No log	10.96	33	3.9336	0.185	0.8746	6.7953	0.185	0.0488	0.2533	0.7685
No log	11.96	36	3.9298	0.185	0.8755	6.9469	0.185	0.0488	0.2679	0.7659
No log	12.96	39	3.9253	0.185	0.8756	6.9654	0.185	0.0488	0.2591	0.7640
No log	13.96	42	3.9194	0.185	0.8750	6.9522	0.185	0.0488	0.2681	0.7604
No log	14.96	45	3.9128	0.185	0.8744	6.9200	0.185	0.0488	0.2611	0.7617
No log	15.96	48	3.9074	0.185	0.8733	6.8369	0.185	0.0488	0.2611	0.7600
No log	16.96	51	3.9041	0.185	0.8726	6.8278	0.185	0.0488	0.2558	0.7566
No log	17.96	54	3.9025	0.185	0.8719	6.7039	0.185	0.0488	0.2588	0.7510
No log	18.96	57	3.9012	0.185	0.8717	6.6384	0.185	0.0488	0.2580	0.7484
No log	19.96	60	3.8987	0.185	0.8712	6.6323	0.185	0.0488	0.2612	0.7450
No log	20.96	63	3.8971	0.185	0.8712	6.6319	0.185	0.0488	0.2615	0.7443
No log	21.96	66	3.8956	0.185	0.8710	6.6323	0.185	0.0488	0.2659	0.7439
No log	22.96	69	3.8945	0.185	0.8708	6.6307	0.185	0.0488	0.2569	0.7436
No log	23.96	72	3.8940	0.185	0.8708	6.6295	0.185	0.0488	0.2526	0.7434
No log	24.96	75	3.8936	0.185	0.8707	6.6284	0.185	0.0488	0.2527	0.7434

Framework versions

• Transformers 4.26.1
• Pytorch 1.13.1.post200
• Datasets 2.9.0
• Tokenizers 0.13.2