videomae-large-finetuned-kinetics-finetuned-rwf2000-epochs8-batch8-kl-torch2

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

• Loss: 0.6146
• Accuracy: 0.7212

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: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• training_steps: 3200
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
0.361	0.06	200	0.2425	0.895
0.3449	1.06	400	0.6639	0.68
0.2435	2.06	600	0.9180	0.6663
0.2001	3.06	800	0.5656	0.7662
0.1405	4.06	1000	0.3859	0.86
0.1845	5.06	1200	0.3825	0.8675
0.1586	6.06	1400	1.4446	0.6687
0.2013	7.06	1600	0.4730	0.8562
0.2113	8.06	1800	0.3328	0.8862
0.245	9.06	2000	0.3519	0.8938
0.1767	10.06	2200	0.4004	0.895
0.1688	11.06	2400	0.6468	0.86
0.2823	12.06	2600	0.6006	0.8575
0.0928	13.06	2800	0.5516	0.875
0.0079	14.06	3000	0.5855	0.87
0.0325	15.06	3200	0.4921	0.8925

Framework versions

• Transformers 4.27.4
• Pytorch 2.0.0+cu117
• Datasets 2.11.0
• Tokenizers 0.13.2