dit-small_tobacco3482_kd_CEKD_t1.5_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: 2.5836
• Accuracy: 0.185
• Brier Loss: 0.8652
• Nll: 6.4546
• F1 Micro: 0.185
• F1 Macro: 0.0488
• Ece: 0.2424
• Aurc: 0.7342

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	2.8093	0.06	0.9041	9.2868	0.06	0.0114	0.1752	0.9033
No log	1.96	6	2.7245	0.18	0.8884	6.2166	0.18	0.0305	0.2292	0.8036
No log	2.96	9	2.6443	0.18	0.8760	6.9627	0.18	0.0305	0.2437	0.8179
No log	3.96	12	2.6356	0.185	0.8785	6.9306	0.185	0.0488	0.2534	0.7877
No log	4.96	15	2.6338	0.185	0.8768	6.8870	0.185	0.0488	0.2605	0.7787
No log	5.96	18	2.6325	0.185	0.8740	6.2086	0.185	0.0490	0.2453	0.7699
No log	6.96	21	2.6322	0.185	0.8721	5.9554	0.185	0.0488	0.2474	0.7629
No log	7.96	24	2.6293	0.185	0.8712	5.9359	0.185	0.0488	0.2550	0.7576
No log	8.96	27	2.6221	0.185	0.8701	5.9468	0.185	0.0488	0.2436	0.7536
No log	9.96	30	2.6171	0.185	0.8697	6.6875	0.185	0.0488	0.2497	0.7541
No log	10.96	33	2.6126	0.185	0.8697	6.7549	0.185	0.0488	0.2512	0.7517
No log	11.96	36	2.6084	0.185	0.8697	6.7827	0.185	0.0488	0.2476	0.7489
No log	12.96	39	2.6037	0.185	0.8692	6.7652	0.185	0.0488	0.2557	0.7476
No log	13.96	42	2.5986	0.185	0.8683	6.6847	0.185	0.0488	0.2513	0.7446
No log	14.96	45	2.5940	0.185	0.8676	6.6600	0.185	0.0488	0.2572	0.7447
No log	15.96	48	2.5910	0.185	0.8669	6.6410	0.185	0.0488	0.2448	0.7424
No log	16.96	51	2.5897	0.185	0.8667	6.6371	0.185	0.0488	0.2402	0.7402
No log	17.96	54	2.5898	0.185	0.8664	6.5096	0.185	0.0488	0.2549	0.7371
No log	18.96	57	2.5897	0.185	0.8664	6.5160	0.185	0.0488	0.2504	0.7363
No log	19.96	60	2.5877	0.185	0.8660	6.4661	0.185	0.0488	0.2416	0.7346
No log	20.96	63	2.5865	0.185	0.8658	6.4833	0.185	0.0488	0.2459	0.7347
No log	21.96	66	2.5852	0.185	0.8655	6.4690	0.185	0.0488	0.2460	0.7343
No log	22.96	69	2.5843	0.185	0.8654	6.4625	0.185	0.0488	0.2461	0.7340
No log	23.96	72	2.5838	0.185	0.8653	6.4568	0.185	0.0488	0.2424	0.7342
No log	24.96	75	2.5836	0.185	0.8652	6.4546	0.185	0.0488	0.2424	0.7342

Framework versions

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