videomae-base-finetuned-basketball-subset-20epochs

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: 2.8785
• Accuracy: 0.1972

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: 1
• eval_batch_size: 1
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• training_steps: 4060

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
1.2525	0.05	200	0.7720	0.52
0.8649	1.05	400	0.7721	0.48
1.0703	2.05	600	1.3605	0.52
0.606	3.05	800	1.0668	0.6
2.0221	4.05	1000	1.1741	0.56
1.2916	5.05	1200	1.4747	0.52
1.4861	6.05	1400	1.1454	0.6
1.3012	7.05	1600	1.6105	0.56
1.3327	8.05	1800	1.2343	0.52
2.077	9.05	2000	1.3243	0.6
1.2349	10.05	2200	1.2044	0.6
1.005	11.05	2400	1.6417	0.52
1.1622	12.05	2600	1.3058	0.56
0.8031	13.05	2800	0.6776	0.48
0.8588	14.05	3000	1.1644	0.64
0.8451	15.05	3200	0.8491	0.64
1.1336	16.05	3400	1.0237	0.6
1.5719	17.05	3600	1.0391	0.64
0.4892	18.05	3800	0.9995	0.64
1.2092	19.05	4000	0.9802	0.56
0.9784	20.01	4060	0.9771	0.56

Framework versions

• Transformers 4.28.1
• Pytorch 2.0.0+cu118
• Datasets 2.11.0
• Tokenizers 0.13.3