jialicheng/ucf101_videomae-large

ucf101_42

This model is a fine-tuned version of MCG-NJU/videomae-large on the ucf101 dataset. It achieves the following results on the evaluation set:

• Loss: 0.3185
• Accuracy: 0.9345

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: 4
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
No log	1.0	298	0.5806	0.8423
No log	2.0	596	0.5192	0.8653
No log	3.0	894	0.4903	0.8814
0.5923	4.0	1192	0.4623	0.8901
0.5923	5.0	1490	0.3949	0.9005
0.5923	6.0	1788	0.4748	0.8844
0.0581	7.0	2086	0.4877	0.8820
0.0581	8.0	2385	0.3976	0.9131
0.0581	9.0	2683	0.3824	0.9116
0.0581	10.0	2981	0.3553	0.9171
0.0221	11.0	3279	0.3557	0.9229
0.0221	12.0	3577	0.3619	0.9258
0.0221	13.0	3875	0.3941	0.9214
0.0112	14.0	4173	0.3989	0.9145
0.0112	15.0	4471	0.3635	0.9236
0.0112	16.0	4770	0.3418	0.9285
0.005	17.0	5068	0.3374	0.9261
0.005	18.0	5366	0.3340	0.9333
0.005	19.0	5664	0.3294	0.9338
0.005	19.99	5960	0.3185	0.9345

Framework versions

• Transformers 4.39.3
• Pytorch 2.2.2+cu118
• Datasets 2.18.0
• Tokenizers 0.15.2