VideoMAE_wlasl_100_200_epochs_longtail

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: 2.9454
• Accuracy: 0.5414

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	Accuracy
18.6587	0.005	180	4.6335	0.0089
18.6163	1.0050	360	4.6142	0.0237
18.5841	2.0050	540	4.6104	0.0118
18.3966	3.0050	721	4.6061	0.0207
18.4731	4.005	901	4.6159	0.0118
18.3719	5.0050	1081	4.6205	0.0237
18.355	6.0050	1261	4.6272	0.0207
18.0586	7.0050	1442	4.6182	0.0266
18.0677	8.005	1622	4.6398	0.0237
17.8698	9.0050	1802	4.6682	0.0207
17.5019	10.0050	1982	4.4023	0.0296
16.5777	11.0050	2163	4.1494	0.0355
15.5504	12.005	2343	3.8834	0.0828
14.2355	13.0050	2523	3.6940	0.1065
12.806	14.0050	2703	3.4313	0.1657
11.2879	15.0050	2884	3.1503	0.2278
9.6452	16.005	3064	2.8827	0.3018
8.1764	17.0050	3244	2.6588	0.3639
6.7153	18.0050	3424	2.4240	0.4142
5.402	19.0050	3605	2.3956	0.3846
4.2777	20.005	3785	2.1764	0.4497
3.3083	21.0050	3965	2.1326	0.4438
2.6866	22.0050	4145	2.0766	0.4615
1.9762	23.0050	4326	1.9066	0.5266
1.4935	24.005	4506	1.9463	0.5355
1.2328	25.0050	4686	2.0531	0.4763
0.992	26.0050	4866	2.0038	0.4970
0.7934	27.0050	5047	2.0482	0.4970
0.5944	28.005	5227	2.1296	0.5059
0.5528	29.0050	5407	2.1365	0.5266
0.3907	30.0050	5587	2.2192	0.5118
0.3661	31.0050	5768	2.1891	0.5296
0.3419	32.005	5948	2.3824	0.4941
0.2103	33.0050	6128	2.2075	0.5385
0.189	34.0050	6308	2.2808	0.5296
0.1627	35.0050	6489	2.3610	0.5237
0.1798	36.005	6669	2.4350	0.5207
0.2012	37.0050	6849	2.4908	0.5
0.1502	38.0050	7029	2.5035	0.5503
0.1678	39.0050	7210	2.3674	0.5828
0.1638	40.005	7390	2.4963	0.5680
0.2638	41.0050	7570	2.5150	0.5325
0.1584	42.0050	7750	2.3931	0.5562
0.1268	43.0050	7931	2.4318	0.5740
0.2002	44.005	8111	2.5972	0.5355
0.2458	45.0050	8291	2.7025	0.5473
0.245	46.0050	8471	2.6003	0.5473
0.3402	47.0050	8652	2.9198	0.4970
0.2087	48.005	8832	2.9450	0.5118
0.1762	49.0050	9012	2.6300	0.5473
0.2306	50.0050	9192	2.8671	0.5118
0.196	51.0050	9373	2.9024	0.5178
0.172	52.005	9553	3.0055	0.5207
0.119	53.0050	9733	2.8886	0.5325
0.3371	54.0050	9913	3.0279	0.5118
0.2315	55.0050	10094	2.9214	0.5680
0.1184	56.005	10274	2.8867	0.5533
0.1617	57.0050	10454	3.2441	0.5059
0.1067	58.0050	10634	2.9521	0.5621
0.1099	59.0050	10815	2.9454	0.5414

Framework versions

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