videomae-base-finetuned-good-gesturePhaseV3

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:

• Accuracy: 0.6939
• Loss: 0.9297
• Accuracy Hold: 0.0
• Accuracy Stroke: 0.0
• Accuracy Recovery: 0.0
• Accuracy Preparation: 0.8961
• Accuracy Unknown: 0.9429

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: 1e-05
• train_batch_size: 8
• eval_batch_size: 8
• 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: 690

Training results

Training Loss	Epoch	Step	Accuracy	Validation Loss	Accuracy Hold	Accuracy Stroke	Accuracy Recovery	Accuracy Preparation	Accuracy Unknown
1.0859	0.1014	70	0.5597	1.2821	0.0	0.0	0.0	1.0	0.0
1.1816	1.1014	140	0.5597	1.2569	0.0	0.0	0.0	1.0	0.0
0.9776	2.1014	210	0.5597	1.1835	0.0	0.0	0.0	1.0	0.0
0.9867	3.1014	280	0.6119	1.0721	0.0	0.0	0.0	0.96	0.3846
1.0732	4.1014	350	0.6045	1.1086	0.0	0.0	0.0	0.7333	1.0
0.9688	5.1014	420	0.6866	0.9288	0.0	0.0	0.0	0.9333	0.8462
0.7733	6.1014	490	0.6567	0.9749	0.0	0.0	0.0	0.8267	1.0
0.8798	7.1014	560	0.6866	0.8742	0.0	0.0	0.0	0.92	0.8846
0.7771	8.1014	630	0.7015	0.8452	0.0	0.0	0.0	0.92	0.9615
1.0448	9.0870	690	0.7239	0.8357	0.0	0.0	0.0	0.96	0.9615

Framework versions

• Transformers 4.41.1
• Pytorch 2.3.0+cu121
• Datasets 2.19.1
• Tokenizers 0.19.1