dit-small_tobacco3482_kd_CEKD_t1.5_a0.9

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: 2.2890
• Accuracy: 0.19
• Brier Loss: 0.8648
• Nll: 6.4150
• F1 Micro: 0.19
• F1 Macro: 0.0641
• Ece: 0.2450
• Aurc: 0.7332

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	2.4806	0.06	0.9041	9.2838	0.06	0.0114	0.1750	0.9034
No log	1.96	6	2.4041	0.18	0.8884	6.3227	0.18	0.0305	0.2317	0.8027
No log	2.96	9	2.3381	0.18	0.8760	6.9952	0.18	0.0305	0.2424	0.8118
No log	3.96	12	2.3362	0.185	0.8771	6.9040	0.185	0.0488	0.2544	0.7841
No log	4.96	15	2.3345	0.185	0.8747	6.8515	0.185	0.0488	0.2476	0.7768
No log	5.96	18	2.3339	0.185	0.8725	6.0111	0.185	0.0490	0.2457	0.7670
No log	6.96	21	2.3348	0.185	0.8718	5.9199	0.185	0.0488	0.2328	0.7596
No log	7.96	24	2.3310	0.185	0.8711	5.9008	0.185	0.0488	0.2443	0.7536
No log	8.96	27	2.3231	0.185	0.8699	5.8793	0.185	0.0488	0.2337	0.7516
No log	9.96	30	2.3181	0.185	0.8694	6.6980	0.185	0.0488	0.2507	0.7500
No log	10.96	33	2.3139	0.185	0.8692	6.7350	0.185	0.0488	0.2481	0.7488
No log	11.96	36	2.3099	0.185	0.8690	6.7557	0.185	0.0488	0.2484	0.7463
No log	12.96	39	2.3057	0.185	0.8684	6.6765	0.185	0.0488	0.2598	0.7441
No log	13.96	42	2.3014	0.185	0.8676	6.6313	0.185	0.0488	0.2478	0.7420
No log	14.96	45	2.2978	0.185	0.8669	6.6142	0.185	0.0488	0.2496	0.7412
No log	15.96	48	2.2955	0.185	0.8664	6.5990	0.185	0.0488	0.2379	0.7399
No log	16.96	51	2.2947	0.185	0.8662	6.4895	0.185	0.0488	0.2452	0.7375
No log	17.96	54	2.2949	0.185	0.8661	6.4730	0.185	0.0488	0.2438	0.7354
No log	18.96	57	2.2949	0.185	0.8661	6.4244	0.185	0.0488	0.2435	0.7356
No log	19.96	60	2.2930	0.185	0.8657	6.3676	0.185	0.0490	0.2389	0.7341
No log	20.96	63	2.2918	0.19	0.8654	6.4233	0.19	0.0641	0.2446	0.7336
No log	21.96	66	2.2905	0.19	0.8651	6.4742	0.19	0.0641	0.2485	0.7334
No log	22.96	69	2.2897	0.19	0.8649	6.4243	0.19	0.0641	0.2448	0.7332
No log	23.96	72	2.2893	0.19	0.8648	6.4174	0.19	0.0641	0.2450	0.7332
No log	24.96	75	2.2890	0.19	0.8648	6.4150	0.19	0.0641	0.2450	0.7332

Framework versions

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