Amoros_Beaugosse_test-large-2025_05_20_57422-bs16_freeze

This model is a fine-tuned version of facebook/dinov2-large on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.0987
• F1 Micro: 0.5482
• F1 Macro: 0.3454
• Accuracy: 0.4313
• Learning Rate: 1e-05

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: 0.001
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 100
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	F1 Micro	F1 Macro	Accuracy	Rate
No log	1.0	195	0.8014	0.1708	0.1167	0.1568	0.001
No log	2.0	390	0.1238	0.3956	0.1496	0.2677	0.001
0.2531	3.0	585	0.1158	0.4517	0.1890	0.3270	0.001
0.2531	4.0	780	0.1146	0.4248	0.1657	0.2945	0.001
0.2531	5.0	975	0.1105	0.4866	0.2335	0.3652	0.001
0.1272	6.0	1170	0.1105	0.4681	0.2026	0.3442	0.001
0.1272	7.0	1365	0.1122	0.4576	0.2113	0.3308	0.001
0.1158	8.0	1560	0.1062	0.4813	0.2474	0.3461	0.001
0.1158	9.0	1755	0.1058	0.4863	0.2428	0.3537	0.001
0.1158	10.0	1950	0.1075	0.4813	0.2409	0.3480	0.001
0.1126	11.0	2145	0.1058	0.4942	0.2747	0.3595	0.001
0.1126	12.0	2340	0.1065	0.4944	0.2533	0.3738	0.001
0.1098	13.0	2535	0.1071	0.4823	0.2584	0.3614	0.001
0.1098	14.0	2730	0.1086	0.5022	0.2807	0.3776	0.001
0.1098	15.0	2925	0.1056	0.5032	0.2649	0.3767	0.001
0.108	16.0	3120	0.1077	0.5106	0.2838	0.3958	0.001
0.108	17.0	3315	0.1072	0.4892	0.2847	0.3709	0.001
0.1064	18.0	3510	0.1057	0.4962	0.2752	0.3709	0.001
0.1064	19.0	3705	0.1053	0.4917	0.2690	0.3671	0.001
0.1064	20.0	3900	0.1052	0.5016	0.2853	0.3738	0.001
0.1047	21.0	4095	0.1064	0.5085	0.2684	0.3728	0.001
0.1047	22.0	4290	0.1057	0.5067	0.3060	0.3767	0.001
0.1047	23.0	4485	0.1059	0.5094	0.3023	0.3862	0.001
0.1049	24.0	4680	0.1055	0.5167	0.3091	0.3910	0.001
0.1049	25.0	4875	0.1059	0.4856	0.3083	0.3614	0.001
0.1037	26.0	5070	0.1073	0.4931	0.2853	0.3738	0.001
0.1037	27.0	5265	0.1006	0.5214	0.3116	0.3948	0.0001
0.1037	28.0	5460	0.1003	0.5369	0.3350	0.4092	0.0001
0.0968	29.0	5655	0.1000	0.5425	0.3217	0.4197	0.0001
0.0968	30.0	5850	0.0993	0.5555	0.3540	0.4331	0.0001
0.0932	31.0	6045	0.0993	0.5562	0.3749	0.4359	0.0001
0.0932	32.0	6240	0.0990	0.5597	0.3724	0.4369	0.0001
0.0932	33.0	6435	0.0991	0.5521	0.3543	0.4293	0.0001
0.0899	34.0	6630	0.0986	0.5636	0.3742	0.4446	0.0001
0.0899	35.0	6825	0.0990	0.5634	0.3743	0.4455	0.0001
0.0882	36.0	7020	0.0997	0.5573	0.3655	0.4407	0.0001
0.0882	37.0	7215	0.1018	0.5635	0.3725	0.4493	0.0001
0.0882	38.0	7410	0.0988	0.5694	0.3782	0.4541	0.0001
0.0846	39.0	7605	0.0991	0.5684	0.3795	0.4541	0.0001
0.0846	40.0	7800	0.0994	0.5692	0.3688	0.4570	0.0001
0.0846	41.0	7995	0.0995	0.5711	0.3758	0.4551	1e-05
0.0832	42.0	8190	0.0991	0.5714	0.3770	0.4608	1e-05
0.0832	43.0	8385	0.0993	0.5753	0.3807	0.4598	1e-05
0.0824	44.0	8580	0.0988	0.5640	0.3753	0.4503	1e-05

Framework versions

• Transformers 4.48.0
• Pytorch 2.6.0+cu118
• Datasets 3.0.2
• Tokenizers 0.21.1