VideoMAE_WLASL_100_200_epochs_p20_SR_8_kinetics

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

• Loss: 1.6213
• Top 1 Accuracy: 0.6716
• Top 5 Accuracy: 0.8994
• Top 10 Accuracy: 0.9379
• Accuracy: 0.6716
• Precision: 0.7449
• Recall: 0.6716
• F1: 0.6675

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: 36000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Top 1 Accuracy	Top 5 Accuracy	Top 10 Accuracy	Accuracy	Precision	Recall	F1
18.5163	0.005	180	4.6097	0.0089	0.0562	0.1124	0.0089	0.0005	0.0089	0.0008
18.4047	1.0050	360	4.5832	0.0178	0.0740	0.1361	0.0178	0.0025	0.0178	0.0043
18.096	2.0050	540	4.4750	0.0444	0.1420	0.2219	0.0444	0.0259	0.0444	0.0210
17.0336	3.0050	721	4.1906	0.0917	0.3166	0.4556	0.0917	0.0847	0.0917	0.0612
15.1957	4.005	901	3.8256	0.2633	0.5207	0.6213	0.2633	0.2334	0.2633	0.1964
13.6256	5.0050	1081	3.4320	0.3254	0.6479	0.7574	0.3254	0.3160	0.3254	0.2719
11.5827	6.0050	1261	3.0442	0.4112	0.7337	0.8609	0.4112	0.3981	0.4112	0.3537
9.4052	7.0050	1442	2.6789	0.5266	0.8136	0.8876	0.5266	0.5014	0.5266	0.4744
7.3195	8.005	1622	2.3768	0.5562	0.8639	0.9172	0.5562	0.6125	0.5562	0.5358
5.6096	9.0050	1802	2.0526	0.6331	0.8728	0.9349	0.6331	0.6811	0.6331	0.6143
4.1271	10.0050	1982	1.8377	0.6746	0.8846	0.9467	0.6746	0.7145	0.6746	0.6560
2.8909	11.0050	2163	1.6035	0.6864	0.9053	0.9527	0.6864	0.7437	0.6864	0.6756
2.11	12.005	2343	1.4429	0.6893	0.9053	0.9497	0.6893	0.7392	0.6893	0.6790
1.3243	13.0050	2523	1.2918	0.7130	0.9290	0.9704	0.7130	0.7461	0.7130	0.6986
0.9066	14.0050	2703	1.2568	0.7041	0.9349	0.9734	0.7041	0.7495	0.7041	0.6946
0.5573	15.0050	2884	1.1904	0.7101	0.9290	0.9675	0.7071	0.7494	0.7071	0.7009
0.4602	16.005	3064	1.1545	0.7337	0.9231	0.9556	0.7337	0.7812	0.7337	0.7277
0.2747	17.0050	3244	1.2449	0.6805	0.9142	0.9497	0.6805	0.7267	0.6805	0.6706
0.241	18.0050	3424	1.1410	0.6953	0.9290	0.9645	0.6953	0.7501	0.6953	0.6932
0.2258	19.0050	3605	1.0789	0.7130	0.9201	0.9556	0.7130	0.7400	0.7130	0.6938
0.1156	20.005	3785	1.1841	0.7130	0.9172	0.9408	0.7130	0.7546	0.7130	0.7053
0.1339	21.0050	3965	1.1466	0.7130	0.9053	0.9556	0.7130	0.7382	0.7130	0.6982
0.0827	22.0050	4145	1.1599	0.7189	0.9231	0.9615	0.7219	0.7461	0.7219	0.7062
0.0987	23.0050	4326	1.2831	0.7160	0.9201	0.9615	0.7160	0.7579	0.7160	0.7049
0.1273	24.005	4506	1.2927	0.7041	0.9172	0.9586	0.7041	0.7634	0.7041	0.7014
0.086	25.0050	4686	1.3708	0.7041	0.8994	0.9408	0.7041	0.7368	0.7041	0.6908
0.1436	26.0050	4866	1.2470	0.7189	0.9172	0.9586	0.7189	0.7526	0.7189	0.7097
0.0837	27.0050	5047	1.3399	0.7041	0.9201	0.9615	0.7041	0.7554	0.7041	0.6993
0.0605	28.005	5227	1.3397	0.7219	0.9172	0.9586	0.7219	0.7781	0.7219	0.7160
0.0508	29.0050	5407	1.2904	0.7130	0.9083	0.9645	0.7130	0.7471	0.7130	0.7022
0.1116	30.0050	5587	1.5462	0.6746	0.9112	0.9556	0.6746	0.7478	0.6746	0.6694
0.0343	31.0050	5768	1.4960	0.7012	0.9112	0.9586	0.7012	0.7383	0.7012	0.6832
0.0886	32.005	5948	1.5807	0.6716	0.8876	0.9290	0.6716	0.6948	0.6716	0.6494
0.0852	33.0050	6128	1.4407	0.6953	0.9201	0.9527	0.6953	0.7420	0.6953	0.6886
0.111	34.0050	6308	1.6302	0.6598	0.8994	0.9379	0.6598	0.7054	0.6598	0.6543
0.1465	35.0050	6489	1.4035	0.7071	0.8905	0.9556	0.7071	0.7528	0.7071	0.7023
0.1215	36.005	6669	1.6213	0.6716	0.8994	0.9379	0.6716	0.7449	0.6716	0.6675

Framework versions

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