videomae-base-finetuned-rwf2000-subset___v3

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: 0.4549
• Accuracy: 0.86

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: 4.5e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• 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: 2000

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
0.5109	0.05	100	0.5774	0.6637
0.3791	1.05	200	0.8587	0.6625
0.3991	2.05	300	0.8032	0.66
0.3369	3.05	400	0.4175	0.8125
0.299	4.05	500	0.6348	0.7338
0.2876	5.05	600	0.7832	0.7075
0.4035	6.05	700	0.4888	0.7863
0.2898	7.05	800	0.4414	0.8075
0.2241	8.05	900	0.3882	0.845
0.2225	9.05	1000	0.3793	0.8512
0.2782	10.05	1100	0.5330	0.8013
0.2424	11.05	1200	0.5796	0.7887
0.1123	12.05	1300	0.5412	0.8263
0.2679	13.05	1400	0.6313	0.8163
0.1511	14.05	1500	0.8940	0.7525
0.1491	15.05	1600	0.8924	0.7625
0.0957	16.05	1700	0.8371	0.8
0.1174	17.05	1800	1.0179	0.7625
0.1375	18.05	1900	0.8003	0.8
0.1027	19.05	2000	0.9485	0.7837

Framework versions

• Transformers 4.46.2
• Pytorch 1.13.1+cu117
• Datasets 3.1.0
• Tokenizers 0.20.3