VideoMAE_BdSLW60_FrameRate_Corrected_with_Augment_20_epoch

This model is a fine-tuned version of MCG-NJU/videomae-base-finetuned-kinetics on BdSLW60 dataset. It achieves the following results on the evaluation set (Val):

• Loss: 0.0299
• Accuracy: 0.9965
• Precision: 0.9967
• Recall: 0.9965
• F1: 0.9965

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: 17940
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Precision	Recall	F1
2.4565	0.05	897	1.8706	0.5988	0.6186	0.5988	0.5527
0.5992	1.0500	1795	0.5877	0.8447	0.8821	0.8447	0.8382
0.1653	2.0500	2693	0.2113	0.9435	0.9539	0.9435	0.9428
0.153	3.0500	3591	0.1834	0.9541	0.9615	0.9541	0.9543
0.0552	4.05	4488	0.0852	0.9812	0.9834	0.9812	0.9812
0.0803	5.0500	5386	0.0895	0.9753	0.9775	0.9753	0.9753
0.0457	6.0500	6284	0.0984	0.9788	0.9806	0.9788	0.9788
0.0463	7.0500	7182	0.1675	0.9612	0.9658	0.9612	0.9614
0.0039	8.05	8079	0.0545	0.9882	0.9891	0.9882	0.9882
0.0162	9.0500	8977	0.0891	0.9812	0.9827	0.9812	0.9809
0.0326	10.0500	9875	0.0742	0.9894	0.9904	0.9894	0.9895
0.0278	11.0500	10773	0.0832	0.9812	0.9828	0.9812	0.9811
0.0078	12.05	11670	0.0659	0.9882	0.9887	0.9882	0.9882
0.0004	13.0500	12568	0.0693	0.9882	0.9889	0.9882	0.9882
0.0024	14.0500	13466	0.0517	0.9941	0.9946	0.9941	0.9942
0.0094	15.0500	14364	0.0327	0.9941	0.9944	0.9941	0.9941
0.0001	16.05	15261	0.0342	0.9953	0.9956	0.9953	0.9953
0.0001	17.0500	16159	0.0320	0.9953	0.9956	0.9953	0.9953
0.0025	18.0500	17057	0.0317	0.9953	0.9955	0.9953	0.9953
0.0001	19.0492	17940	0.0299	0.9965	0.9967	0.9965	0.9965

Framework versions

• Transformers 4.44.2
• Pytorch 2.4.1+cu121
• Datasets 2.19.2
• Tokenizers 0.19.1