dit-tiny_tobacco3482_kd_CEKD_t5.0_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: 3.8497
• Accuracy: 0.18
• Brier Loss: 0.8788
• Nll: 6.0432
• F1 Micro: 0.18
• F1 Macro: 0.0305
• Ece: 0.2578
• Aurc: 0.8511

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	4.0678	0.145	0.8999	10.1608	0.145	0.0253	0.2221	0.8466
No log	1.96	6	4.0316	0.145	0.8948	10.5160	0.145	0.0253	0.2239	0.8468
No log	2.96	9	3.9774	0.16	0.8871	8.6333	0.16	0.0524	0.2217	0.8424
No log	3.96	12	3.9325	0.155	0.8813	6.5340	0.155	0.0272	0.2161	0.8837
No log	4.96	15	3.9041	0.155	0.8787	7.1704	0.155	0.0271	0.2296	0.8923
No log	5.96	18	3.8876	0.155	0.8782	8.7334	0.155	0.0277	0.2325	0.8942
No log	6.96	21	3.8766	0.18	0.8785	8.8120	0.18	0.0314	0.2476	0.8555
No log	7.96	24	3.8690	0.18	0.8791	8.8676	0.18	0.0308	0.2643	0.8534
No log	8.96	27	3.8633	0.18	0.8793	8.5299	0.18	0.0306	0.2594	0.8541
No log	9.96	30	3.8601	0.18	0.8796	7.4142	0.18	0.0305	0.2622	0.8548
No log	10.96	33	3.8577	0.18	0.8797	6.6642	0.18	0.0305	0.2720	0.8546
No log	11.96	36	3.8560	0.18	0.8797	6.2862	0.18	0.0305	0.2723	0.8543
No log	12.96	39	3.8547	0.18	0.8796	6.2084	0.18	0.0305	0.2678	0.8541
No log	13.96	42	3.8535	0.18	0.8794	6.1826	0.18	0.0305	0.2631	0.8534
No log	14.96	45	3.8525	0.18	0.8793	6.1744	0.18	0.0305	0.2593	0.8529
No log	15.96	48	3.8516	0.18	0.8792	6.1606	0.18	0.0305	0.2680	0.8527
No log	16.96	51	3.8511	0.18	0.8791	6.1634	0.18	0.0305	0.2724	0.8528
No log	17.96	54	3.8510	0.18	0.8791	6.0971	0.18	0.0305	0.2676	0.8525
No log	18.96	57	3.8508	0.18	0.8790	6.0686	0.18	0.0305	0.2630	0.8522
No log	19.96	60	3.8503	0.18	0.8789	6.0495	0.18	0.0305	0.2581	0.8518
No log	20.96	63	3.8501	0.18	0.8789	6.0918	0.18	0.0305	0.2581	0.8516
No log	21.96	66	3.8499	0.18	0.8788	6.0464	0.18	0.0305	0.2536	0.8516
No log	22.96	69	3.8497	0.18	0.8788	6.0419	0.18	0.0305	0.2535	0.8513
No log	23.96	72	3.8497	0.18	0.8788	6.0432	0.18	0.0305	0.2578	0.8511
No log	24.96	75	3.8497	0.18	0.8788	6.0432	0.18	0.0305	0.2578	0.8511

Framework versions

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