videomae-base-finetuned-soccer-action-recognitionx

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: 0.2575
• Accuracy: 0.9281

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: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• training_steps: 1376

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
No log	0.03	43	1.8151	0.3661
1.7589	1.03	86	0.9133	0.6915
1.7589	2.03	129	0.5026	0.7966
0.5494	3.03	172	0.3937	0.8678
0.5494	4.03	215	0.3943	0.8678
0.2747	5.03	258	0.3876	0.8678
0.2747	6.03	301	0.2803	0.9085
0.1731	7.03	345	0.2414	0.9119
0.1731	8.03	388	0.2434	0.9288
0.1165	9.03	431	0.1560	0.9458
0.1165	10.03	474	0.1894	0.9254
0.07	11.03	517	0.2401	0.9153
0.0417	12.03	560	0.1883	0.9390
0.0417	13.03	603	0.2589	0.9051
0.0362	14.03	646	0.2125	0.9492
0.0362	15.03	690	0.2228	0.9390
0.0348	16.03	733	0.1797	0.9525
0.0348	17.03	776	0.1728	0.9390
0.0129	18.03	819	0.2253	0.9254
0.0129	19.03	862	0.1983	0.9356
0.0112	20.03	905	0.2821	0.9220
0.0112	21.03	948	0.2527	0.9356
0.0165	22.03	991	0.2598	0.9288
0.0165	23.03	1035	0.2690	0.9288
0.0056	24.03	1078	0.2817	0.9220
0.0026	25.03	1121	0.2039	0.9424
0.0026	26.03	1164	0.2164	0.9458
0.0033	27.03	1207	0.2063	0.9424
0.0033	28.03	1250	0.1836	0.9525
0.0038	29.03	1293	0.1899	0.9458
0.0038	30.03	1336	0.1815	0.9492
0.0024	31.03	1376	0.1816	0.9525

Framework versions

• Transformers 4.35.0
• Pytorch 2.1.0+cu118
• Datasets 2.14.6
• Tokenizers 0.14.1