dit-tiny_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: 2.3286
• Accuracy: 0.18
• Brier Loss: 0.8742
• Nll: 6.7213
• F1 Micro: 0.18
• F1 Macro: 0.0306
• Ece: 0.2558
• Aurc: 0.8491

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	2.4683	0.145	0.8999	10.1538	0.145	0.0253	0.2220	0.8466
No log	1.96	6	2.4396	0.145	0.8947	10.5704	0.145	0.0253	0.2237	0.8463
No log	2.96	9	2.3985	0.145	0.8869	8.5511	0.145	0.0451	0.2116	0.8036
No log	3.96	12	2.3677	0.21	0.8810	6.5446	0.2100	0.0611	0.2566	0.8335
No log	4.96	15	2.3517	0.155	0.8780	6.8400	0.155	0.0279	0.2309	0.8894
No log	5.96	18	2.3450	0.18	0.8771	8.1897	0.18	0.0313	0.2495	0.8531
No log	6.96	21	2.3407	0.18	0.8767	7.3073	0.18	0.0306	0.2551	0.8513
No log	7.96	24	2.3371	0.18	0.8763	6.9328	0.18	0.0306	0.2501	0.8520
No log	8.96	27	2.3337	0.18	0.8757	6.8828	0.18	0.0306	0.2507	0.8525
No log	9.96	30	2.3321	0.18	0.8753	6.8682	0.18	0.0306	0.2508	0.8524
No log	10.96	33	2.3312	0.18	0.8751	6.7981	0.18	0.0306	0.2462	0.8521
No log	11.96	36	2.3309	0.18	0.8749	6.7375	0.18	0.0306	0.2531	0.8520
No log	12.96	39	2.3307	0.18	0.8748	6.7235	0.18	0.0306	0.2524	0.8518
No log	13.96	42	2.3304	0.18	0.8747	6.7200	0.18	0.0306	0.2482	0.8514
No log	14.96	45	2.3301	0.18	0.8746	6.7201	0.18	0.0306	0.2410	0.8509
No log	15.96	48	2.3298	0.18	0.8746	6.7182	0.18	0.0306	0.2449	0.8505
No log	16.96	51	2.3295	0.18	0.8745	6.7211	0.18	0.0306	0.2412	0.8500
No log	17.96	54	2.3297	0.18	0.8745	6.7201	0.18	0.0306	0.2449	0.8496
No log	18.96	57	2.3296	0.18	0.8745	6.7216	0.18	0.0306	0.2392	0.8494
No log	19.96	60	2.3292	0.18	0.8744	6.7214	0.18	0.0306	0.2371	0.8494
No log	20.96	63	2.3290	0.18	0.8744	6.7222	0.18	0.0306	0.2371	0.8493
No log	21.96	66	2.3288	0.18	0.8743	6.7227	0.18	0.0306	0.2408	0.8494
No log	22.96	69	2.3286	0.18	0.8743	6.7223	0.18	0.0306	0.2558	0.8490
No log	23.96	72	2.3286	0.18	0.8743	6.7218	0.18	0.0306	0.2558	0.8491
No log	24.96	75	2.3286	0.18	0.8742	6.7213	0.18	0.0306	0.2558	0.8491

Framework versions

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