dit-small_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: 3.7912
• Accuracy: 0.185
• Brier Loss: 0.8688
• Nll: 5.6106
• F1 Micro: 0.185
• F1 Macro: 0.0488
• Ece: 0.2524
• Aurc: 0.7391

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	4.0715	0.06	0.9043	8.8976	0.06	0.0114	0.1751	0.9034
No log	1.96	6	3.9774	0.18	0.8893	8.0316	0.18	0.0305	0.2237	0.8040
No log	2.96	9	3.8805	0.18	0.8782	8.6752	0.18	0.0305	0.2566	0.8189
No log	3.96	12	3.8615	0.18	0.8836	8.9177	0.18	0.0305	0.2645	0.8205
No log	4.96	15	3.8624	0.185	0.8844	6.3245	0.185	0.0488	0.2727	0.7889
No log	5.96	18	3.8605	0.185	0.8813	5.1679	0.185	0.0488	0.2558	0.7797
No log	6.96	21	3.8511	0.185	0.8774	5.1770	0.185	0.0488	0.2510	0.7741
No log	7.96	24	3.8410	0.185	0.8751	5.6014	0.185	0.0488	0.2458	0.7699
No log	8.96	27	3.8317	0.185	0.8733	5.9766	0.185	0.0488	0.2537	0.7681
No log	9.96	30	3.8259	0.185	0.8724	6.0278	0.185	0.0488	0.2473	0.7689
No log	10.96	33	3.8226	0.185	0.8724	6.8070	0.185	0.0488	0.2618	0.7671
No log	11.96	36	3.8209	0.185	0.8730	7.6044	0.185	0.0488	0.2539	0.7643
No log	12.96	39	3.8187	0.185	0.8730	8.1654	0.185	0.0488	0.2542	0.7612
No log	13.96	42	3.8147	0.185	0.8725	7.1073	0.185	0.0488	0.2542	0.7566
No log	14.96	45	3.8096	0.185	0.8720	6.3875	0.185	0.0488	0.2565	0.7566
No log	15.96	48	3.8052	0.185	0.8712	6.0256	0.185	0.0488	0.2518	0.7524
No log	16.96	51	3.8022	0.185	0.8707	5.7809	0.185	0.0488	0.2558	0.7485
No log	17.96	54	3.8008	0.185	0.8701	5.6835	0.185	0.0488	0.2496	0.7442
No log	18.96	57	3.7992	0.185	0.8700	5.3867	0.185	0.0488	0.2490	0.7421
No log	19.96	60	3.7965	0.185	0.8694	5.4928	0.185	0.0488	0.2478	0.7406
No log	20.96	63	3.7948	0.185	0.8693	5.5527	0.185	0.0488	0.2481	0.7405
No log	21.96	66	3.7932	0.185	0.8691	5.5585	0.185	0.0488	0.2564	0.7396
No log	22.96	69	3.7921	0.185	0.8689	5.5607	0.185	0.0488	0.2479	0.7391
No log	23.96	72	3.7915	0.185	0.8688	5.6116	0.185	0.0488	0.2523	0.7390
No log	24.96	75	3.7912	0.185	0.8688	5.6106	0.185	0.0488	0.2524	0.7391

Framework versions

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