dit-small_tobacco3482_kd_CEKD_t2.5_a0.7

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: 3.1993
• Accuracy: 0.185
• Brier Loss: 0.8672
• Nll: 6.5703
• F1 Micro: 0.185
• F1 Macro: 0.0488
• Ece: 0.2594
• Aurc: 0.7367

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	3.4684	0.06	0.9042	9.2910	0.06	0.0114	0.1755	0.9033
No log	1.96	6	3.3741	0.18	0.8886	6.5491	0.18	0.0305	0.2324	0.8055
No log	2.96	9	3.2779	0.18	0.8767	7.2662	0.18	0.0305	0.2493	0.8196
No log	3.96	12	3.2605	0.18	0.8816	7.0963	0.18	0.0305	0.2628	0.8140
No log	4.96	15	3.2592	0.185	0.8814	6.9350	0.185	0.0488	0.2584	0.7850
No log	5.96	18	3.2576	0.185	0.8782	6.3113	0.185	0.0488	0.2561	0.7731
No log	6.96	21	3.2540	0.185	0.8747	6.0058	0.185	0.0488	0.2446	0.7705
No log	7.96	24	3.2500	0.185	0.8731	5.9849	0.185	0.0488	0.2442	0.7669
No log	8.96	27	3.2430	0.185	0.8717	5.9785	0.185	0.0488	0.2483	0.7626
No log	9.96	30	3.2377	0.185	0.8711	6.2837	0.185	0.0488	0.2462	0.7609
No log	10.96	33	3.2332	0.185	0.8713	6.8641	0.185	0.0488	0.2560	0.7601
No log	11.96	36	3.2293	0.185	0.8719	6.8631	0.185	0.0488	0.2523	0.7587
No log	12.96	39	3.2246	0.185	0.8717	6.8535	0.185	0.0488	0.2526	0.7558
No log	13.96	42	3.2190	0.185	0.8709	6.8177	0.185	0.0488	0.2565	0.7533
No log	14.96	45	3.2134	0.185	0.8700	6.7894	0.185	0.0488	0.2630	0.7533
No log	15.96	48	3.2091	0.185	0.8691	6.7672	0.185	0.0488	0.2585	0.7500
No log	16.96	51	3.2069	0.185	0.8687	6.6512	0.185	0.0488	0.2536	0.7466
No log	17.96	54	3.2063	0.185	0.8682	6.5227	0.185	0.0488	0.2520	0.7429
No log	18.96	57	3.2057	0.185	0.8682	6.5119	0.185	0.0488	0.2514	0.7406
No log	19.96	60	3.2036	0.185	0.8678	6.5674	0.185	0.0488	0.2501	0.7385
No log	20.96	63	3.2023	0.185	0.8677	6.5709	0.185	0.0488	0.2506	0.7385
No log	21.96	66	3.2010	0.185	0.8675	6.5731	0.185	0.0488	0.2631	0.7376
No log	22.96	69	3.2000	0.185	0.8673	6.5723	0.185	0.0488	0.2591	0.7371
No log	23.96	72	3.1996	0.185	0.8673	6.5715	0.185	0.0488	0.2593	0.7368
No log	24.96	75	3.1993	0.185	0.8672	6.5703	0.185	0.0488	0.2594	0.7367

Framework versions

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