VideoMAE_WLASL_100_SR_8

This model is a fine-tuned version of MCG-NJU/videomae-base on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 3.3229
• Accuracy: 0.3964
• Precision: 0.4264
• Recall: 0.3964
• F1: 0.3811

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: 5e-05
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 8
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• training_steps: 9000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Precision	Recall	F1
18.6069	0.02	180	4.6399	0.0	0.0	0.0	0.0
18.5558	1.0200	360	4.6268	0.0148	0.0003	0.0148	0.0006
18.5458	2.0199	540	4.6364	0.0089	0.0003	0.0089	0.0005
18.5593	3.0200	721	4.6385	0.0148	0.0005	0.0148	0.0010
18.5296	4.02	901	4.6196	0.0207	0.0012	0.0207	0.0020
18.355	5.0200	1081	4.6494	0.0178	0.0006	0.0178	0.0011
18.2567	6.0199	1261	4.6268	0.0148	0.0017	0.0148	0.0030
18.0477	7.0200	1442	4.6423	0.0237	0.0040	0.0237	0.0065
17.9032	8.02	1622	4.6046	0.0207	0.0037	0.0207	0.0061
17.6066	9.0200	1802	4.5116	0.0385	0.0074	0.0385	0.0116
16.848	10.0199	1982	4.3914	0.0296	0.0097	0.0296	0.0134
16.2226	11.0200	2163	4.3536	0.0325	0.0219	0.0325	0.0206
15.4599	12.02	2343	4.3156	0.0296	0.0190	0.0296	0.0187
14.6418	13.0200	2523	4.1732	0.0621	0.0318	0.0621	0.0404
13.7576	14.0199	2703	4.0590	0.0947	0.0599	0.0947	0.0654
12.8445	15.0200	2884	3.9267	0.0976	0.0782	0.0976	0.0768
11.3374	16.02	3064	3.8400	0.1124	0.0884	0.1124	0.0803
10.2244	17.0200	3244	3.6886	0.1538	0.1493	0.1538	0.1307
8.5388	18.0199	3424	3.5762	0.1864	0.1925	0.1864	0.1619
7.1906	19.0200	3605	3.4168	0.2130	0.2111	0.2130	0.1949
5.9014	20.02	3785	3.2546	0.2367	0.2693	0.2367	0.2276
4.5854	21.0200	3965	3.1853	0.2574	0.2864	0.2574	0.2479
3.5423	22.0199	4145	3.0606	0.2899	0.3137	0.2899	0.2805
2.6816	23.0200	4326	3.0254	0.3018	0.3183	0.3018	0.2875
1.998	24.02	4506	3.0007	0.2929	0.3343	0.2929	0.2834
1.4233	25.0200	4686	3.0289	0.3225	0.3389	0.3225	0.3061
1.0355	26.0199	4866	2.9665	0.3343	0.3718	0.3343	0.3230
0.7254	27.0200	5047	3.0491	0.3254	0.3580	0.3254	0.3112
0.4656	28.02	5227	3.0303	0.3225	0.3494	0.3225	0.3091
0.3909	29.0200	5407	2.9928	0.3639	0.3901	0.3639	0.3436
0.31	30.0199	5587	3.0805	0.3195	0.3434	0.3195	0.3025
0.179	31.0200	5768	3.0511	0.3491	0.3915	0.3491	0.3349
0.1252	32.02	5948	3.1101	0.3757	0.3966	0.3757	0.3570
0.0966	33.0200	6128	3.1468	0.3846	0.4160	0.3846	0.3691
0.0747	34.0199	6308	3.1351	0.3787	0.4003	0.3787	0.3617
0.0431	35.0200	6489	3.2090	0.3846	0.3979	0.3846	0.3631
0.1095	36.02	6669	3.2229	0.3728	0.3748	0.3728	0.3474
0.0462	37.0200	6849	3.2372	0.3905	0.4124	0.3905	0.3685
0.064	38.0199	7029	3.2563	0.3787	0.4024	0.3787	0.3597
0.125	39.0200	7210	3.2660	0.3817	0.4034	0.3817	0.3614
0.0219	40.02	7390	3.2760	0.3846	0.4123	0.3846	0.3677
0.0767	41.0200	7570	3.2750	0.3698	0.3934	0.3698	0.3528
0.0546	42.0199	7750	3.3079	0.3846	0.4085	0.3846	0.3676
0.0341	43.0200	7931	3.2994	0.3817	0.4116	0.3817	0.3672
0.0252	44.02	8111	3.3123	0.3757	0.3887	0.3757	0.3556
0.0623	45.0200	8291	3.3223	0.3876	0.4077	0.3876	0.3709
0.0439	46.0199	8471	3.3238	0.3935	0.4276	0.3935	0.3772
0.0471	47.0200	8652	3.3215	0.3935	0.4194	0.3935	0.3778
0.047	48.02	8832	3.3259	0.3935	0.4229	0.3935	0.3784
0.045	49.0186	9000	3.3229	0.3964	0.4264	0.3964	0.3811

Framework versions

• Transformers 4.46.1
• Pytorch 2.5.1+cu124
• Datasets 3.1.0
• Tokenizers 0.20.1