resnet101-base_tobacco-cnn_tobacco3482_kd_NKD_t1.0_g1.5

This model is a fine-tuned version of bdpc/resnet101-base_tobacco on the None dataset. It achieves the following results on the evaluation set:

• Loss: 3.2266
• Accuracy: 0.385
• Brier Loss: 0.7374
• Nll: 4.0859
• F1 Micro: 0.3850
• F1 Macro: 0.2652
• Ece: 0.2858
• Aurc: 0.4261

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: 0.0001
• train_batch_size: 256
• eval_batch_size: 256
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Brier Loss	Nll	F1 Micro	F1 Macro	Ece	Aurc
No log	1.0	4	3.9610	0.05	0.9004	9.1922	0.0500	0.0100	0.1457	0.9423
No log	2.0	8	3.8752	0.155	0.8934	8.5838	0.155	0.0268	0.2356	0.9630
No log	3.0	12	3.8665	0.155	0.9261	8.8645	0.155	0.0268	0.3113	0.7263
No log	4.0	16	5.0552	0.155	1.3190	8.8736	0.155	0.0268	0.6667	0.6226
No log	5.0	20	4.9755	0.155	1.2873	8.9603	0.155	0.0270	0.6315	0.6033
No log	6.0	24	4.6069	0.155	1.1443	7.0637	0.155	0.0301	0.5057	0.6065
No log	7.0	28	3.7058	0.22	0.9193	6.6528	0.22	0.0685	0.3454	0.5737
No log	8.0	32	3.3000	0.25	0.8140	6.8642	0.25	0.1011	0.2638	0.5377
No log	9.0	36	3.3805	0.195	0.8768	6.5108	0.195	0.0779	0.2955	0.7532
No log	10.0	40	3.4626	0.2	0.8985	6.3933	0.2000	0.0745	0.3154	0.7384
No log	11.0	44	3.2088	0.32	0.7621	6.0433	0.32	0.1695	0.2375	0.4457
No log	12.0	48	3.4543	0.22	0.8720	6.1413	0.22	0.1065	0.3144	0.7026
No log	13.0	52	3.5300	0.225	0.8684	7.0938	0.225	0.1182	0.2747	0.7110
No log	14.0	56	3.5981	0.215	0.8821	7.5146	0.2150	0.0978	0.3047	0.7351
No log	15.0	60	3.5641	0.23	0.8895	7.7554	0.23	0.0944	0.2985	0.7568
No log	16.0	64	3.5853	0.235	0.8698	6.6949	0.235	0.1292	0.2634	0.6518
No log	17.0	68	3.5539	0.255	0.8597	7.5062	0.255	0.1331	0.2821	0.6332
No log	18.0	72	3.5725	0.265	0.8569	7.4117	0.265	0.1396	0.2708	0.5940
No log	19.0	76	3.5207	0.27	0.8415	6.5482	0.27	0.1542	0.2592	0.5619
No log	20.0	80	3.5360	0.26	0.8573	7.4207	0.26	0.1358	0.2942	0.5949
No log	21.0	84	3.2807	0.345	0.7933	4.8232	0.345	0.2077	0.2903	0.5385
No log	22.0	88	3.1633	0.39	0.7217	4.3843	0.39	0.2417	0.2547	0.3857
No log	23.0	92	3.2159	0.39	0.7463	4.4691	0.39	0.2481	0.2923	0.3756
No log	24.0	96	3.1650	0.375	0.7248	4.4043	0.375	0.2276	0.2433	0.3809
No log	25.0	100	3.2000	0.375	0.7470	4.7004	0.375	0.2473	0.2671	0.4264
No log	26.0	104	3.4356	0.27	0.8326	6.6479	0.27	0.1466	0.2636	0.5640
No log	27.0	108	3.5761	0.285	0.8347	6.5689	0.285	0.1796	0.2537	0.6182
No log	28.0	112	3.5778	0.26	0.8546	7.0753	0.26	0.1380	0.2629	0.5870
No log	29.0	116	3.1280	0.39	0.7075	4.5179	0.39	0.2450	0.2479	0.3759
No log	30.0	120	3.1559	0.37	0.7268	4.3444	0.37	0.2413	0.2588	0.3941
No log	31.0	124	3.1493	0.39	0.7133	4.6188	0.39	0.2305	0.2338	0.3686
No log	32.0	128	3.1287	0.39	0.7015	4.0848	0.39	0.2379	0.2271	0.3655
No log	33.0	132	3.1409	0.395	0.7048	4.0026	0.395	0.2290	0.2210	0.3606
No log	34.0	136	3.1691	0.375	0.7210	4.4086	0.375	0.2215	0.2495	0.3800
No log	35.0	140	3.1529	0.4	0.7117	4.1376	0.4000	0.2487	0.2200	0.3605
No log	36.0	144	3.1088	0.4	0.6989	4.0773	0.4000	0.2645	0.2478	0.3641
No log	37.0	148	3.2158	0.4	0.7230	4.1145	0.4000	0.2603	0.2517	0.3761
No log	38.0	152	3.1351	0.39	0.7064	4.3952	0.39	0.2398	0.2475	0.3606
No log	39.0	156	3.1239	0.395	0.7001	4.0496	0.395	0.2569	0.2364	0.3583
No log	40.0	160	3.1855	0.385	0.7169	4.0634	0.3850	0.2274	0.2467	0.3687
No log	41.0	164	3.1938	0.37	0.7098	3.9505	0.37	0.2146	0.2207	0.3781
No log	42.0	168	3.3495	0.395	0.7438	4.0247	0.395	0.2428	0.2901	0.3973
No log	43.0	172	3.2352	0.395	0.7115	3.9875	0.395	0.2431	0.2651	0.3790
No log	44.0	176	3.2838	0.39	0.7223	3.8867	0.39	0.2246	0.2590	0.3824
No log	45.0	180	3.3175	0.395	0.7304	4.2165	0.395	0.2286	0.2549	0.3811
No log	46.0	184	3.1183	0.395	0.6916	3.9786	0.395	0.2338	0.2345	0.3581
No log	47.0	188	3.1608	0.395	0.7049	3.7245	0.395	0.2580	0.2429	0.3668
No log	48.0	192	3.2144	0.38	0.7316	3.9593	0.38	0.2512	0.2517	0.4202
No log	49.0	196	3.2781	0.365	0.7561	3.9721	0.3650	0.2440	0.2429	0.4654
No log	50.0	200	3.2266	0.385	0.7374	4.0859	0.3850	0.2652	0.2858	0.4261

Framework versions

• Transformers 4.36.0.dev0
• Pytorch 2.2.0.dev20231112+cu118
• Datasets 2.14.5
• Tokenizers 0.14.1