videomae-base-finetuned-soccer-action-recognition

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: 0.2554
• Accuracy: 0.9470

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: 1
• eval_batch_size: 1
• seed: 42
• gradient_accumulation_steps: 8
• 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: 2728

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
1.7115	0.03	85	1.4196	0.4
1.0097	1.03	170	0.7807	0.6759
0.6192	2.03	255	0.7952	0.7034
0.4713	3.03	341	0.6536	0.7931
0.3973	4.03	426	0.3638	0.8690
0.3633	5.03	511	0.3616	0.8966
0.2336	6.03	596	0.4579	0.8966
0.1997	7.03	682	1.5970	0.6069
0.2738	8.03	767	0.4102	0.8690
0.2492	9.03	852	0.7651	0.8345
0.1568	10.03	937	0.8561	0.8138
0.1856	11.03	1023	0.2811	0.9241
0.1296	12.03	1108	0.3444	0.9172
0.0782	13.03	1193	0.3423	0.9241
0.14	14.03	1278	0.3122	0.9241
0.0689	15.03	1364	0.3534	0.9172
0.036	16.03	1449	0.4815	0.9103
0.0695	17.03	1534	0.5698	0.8828
0.0618	18.03	1619	0.3053	0.9310
0.0553	19.03	1705	0.3443	0.9241
0.0301	20.03	1790	0.1427	0.9586
0.0412	21.03	1875	0.5619	0.8690
0.0492	22.03	1960	0.5701	0.8897
0.0171	23.03	2046	0.6377	0.8690
0.0181	24.03	2131	0.5981	0.8828
0.0305	25.03	2216	0.3178	0.9448
0.0393	26.03	2301	0.5434	0.9103
0.0248	27.03	2387	0.4097	0.9241
0.0146	28.03	2472	0.4427	0.9103
0.012	29.03	2557	0.5619	0.9034
0.0065	30.03	2642	0.5384	0.9103
0.009	31.03	2728	0.5014	0.9172

Framework versions

• Transformers 4.34.1
• Pytorch 2.1.0+cu118
• Datasets 2.14.6
• Tokenizers 0.14.1