dit-small_tobacco3482_kd_CEKD_t1.5_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: 2.8753
• Accuracy: 0.185
• Brier Loss: 0.8660
• Nll: 6.5533
• F1 Micro: 0.185
• F1 Macro: 0.0488
• Ece: 0.2451
• Aurc: 0.7363

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	3.1378	0.06	0.9042	9.2898	0.06	0.0114	0.1754	0.9032
No log	1.96	6	3.0447	0.18	0.8884	6.2145	0.18	0.0305	0.2294	0.8048
No log	2.96	9	2.9500	0.18	0.8761	6.9445	0.18	0.0305	0.2447	0.8193
No log	3.96	12	2.9328	0.18	0.8800	6.9512	0.18	0.0305	0.2565	0.8122
No log	4.96	15	2.9305	0.185	0.8793	6.9136	0.185	0.0488	0.2557	0.7823
No log	5.96	18	2.9286	0.185	0.8762	6.7762	0.185	0.0488	0.2533	0.7721
No log	6.96	21	2.9265	0.185	0.8731	5.9902	0.185	0.0488	0.2345	0.7682
No log	7.96	24	2.9240	0.185	0.8718	5.9696	0.185	0.0488	0.2625	0.7621
No log	8.96	27	2.9177	0.185	0.8707	5.9711	0.185	0.0488	0.2463	0.7578
No log	9.96	30	2.9129	0.185	0.8702	6.6932	0.185	0.0488	0.2485	0.7574
No log	10.96	33	2.9082	0.185	0.8704	6.7772	0.185	0.0488	0.2500	0.7560
No log	11.96	36	2.9039	0.185	0.8707	6.8060	0.185	0.0488	0.2464	0.7537
No log	12.96	39	2.8990	0.185	0.8704	6.7988	0.185	0.0488	0.2466	0.7515
No log	13.96	42	2.8933	0.185	0.8696	6.7771	0.185	0.0488	0.2505	0.7479
No log	14.96	45	2.8879	0.185	0.8688	6.7597	0.185	0.0488	0.2523	0.7482
No log	15.96	48	2.8840	0.185	0.8679	6.6825	0.185	0.0488	0.2648	0.7454
No log	16.96	51	2.8822	0.185	0.8676	6.6742	0.185	0.0488	0.2473	0.7425
No log	17.96	54	2.8819	0.185	0.8672	6.5521	0.185	0.0488	0.2479	0.7405
No log	18.96	57	2.8817	0.185	0.8671	6.5498	0.185	0.0488	0.2536	0.7385
No log	19.96	60	2.8797	0.185	0.8667	6.5563	0.185	0.0488	0.2442	0.7371
No log	20.96	63	2.8784	0.185	0.8666	6.6145	0.185	0.0488	0.2528	0.7374
No log	21.96	66	2.8770	0.185	0.8663	6.6084	0.185	0.0488	0.2489	0.7366
No log	22.96	69	2.8760	0.185	0.8662	6.5683	0.185	0.0488	0.2448	0.7360
No log	23.96	72	2.8756	0.185	0.8661	6.5544	0.185	0.0488	0.2450	0.7363
No log	24.96	75	2.8753	0.185	0.8660	6.5533	0.185	0.0488	0.2451	0.7363

Framework versions

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