videomae-base-finetuned-kinetics-fight_22-01-2024

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

• Loss: 0.2265
• Accuracy: 0.9159
• Precision: 0.9507
• Recall: 0.8773

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-07
• train_batch_size: 14
• eval_batch_size: 14
• 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: 15820

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Precision	Recall
0.6066	0.05	792	0.5969	0.7509	0.7817	0.6961
0.429	1.05	1584	0.3898	0.8422	0.8937	0.7767
0.184	2.05	2376	0.2700	0.8859	0.9281	0.8367
0.1911	3.05	3168	0.2280	0.9028	0.9405	0.8601
0.1115	4.05	3960	0.2218	0.9063	0.9436	0.8642
0.1799	5.05	4752	0.2293	0.9090	0.9604	0.8532
0.1282	6.05	5544	0.2265	0.9159	0.9507	0.8773
0.1211	7.05	6336	0.2554	0.9087	0.9562	0.8567
0.076	8.05	7128	0.2738	0.9063	0.9588	0.8491
0.1152	9.05	7920	0.2785	0.9090	0.9541	0.8594
0.0281	10.05	8712	0.2852	0.9118	0.9537	0.8656
0.0806	11.05	9504	0.2994	0.9094	0.9548	0.8594
0.0755	12.05	10296	0.3124	0.9104	0.9556	0.8608
0.0986	13.05	11088	0.3134	0.9114	0.9462	0.8725
0.0222	14.05	11880	0.3241	0.9111	0.9550	0.8629
0.0272	15.05	12672	0.3269	0.9125	0.9503	0.8704
0.0657	16.05	13464	0.3401	0.9097	0.9556	0.8594
0.1083	17.05	14256	0.3424	0.9097	0.9549	0.8601
0.0059	18.05	15048	0.3461	0.9094	0.9555	0.8587
0.0143	19.05	15820	0.3462	0.9094	0.9555	0.8587

Framework versions

• Transformers 4.38.2
• Pytorch 2.1.0+cu121
• Datasets 2.18.0
• Tokenizers 0.15.2