Human-Action-Recognition-VIT-Base-patch16-224

This model is a fine-tuned version of google/vit-base-patch16-224 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.4165
• Accuracy: 0.8782

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: 64
• eval_batch_size: 64
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 256
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
2.6245	0.9873	39	2.0901	0.4429
1.4804	2.0	79	0.7998	0.7841
0.8447	2.9873	118	0.5635	0.8345
0.6764	4.0	158	0.4952	0.8476
0.6202	4.9873	197	0.4849	0.8595
0.5627	6.0	237	0.4604	0.8643
0.5219	6.9873	276	0.4567	0.8663
0.4912	8.0	316	0.4371	0.8718
0.4736	8.9873	355	0.4320	0.8671
0.4347	10.0	395	0.4229	0.8746
0.4274	10.9873	434	0.4211	0.8734
0.4011	12.0	474	0.4234	0.8762
0.3845	12.9873	513	0.4247	0.8742
0.3797	14.0	553	0.4193	0.8746
0.374	14.9873	592	0.4204	0.8766
0.3567	16.0	632	0.4165	0.8782
0.3565	16.9873	671	0.4212	0.8742
0.3375	18.0	711	0.4241	0.875
0.3366	18.9873	750	0.4221	0.875
0.3338	19.7468	780	0.4215	0.8766

Framework versions

• Transformers 4.44.2
• Pytorch 2.5.0+cu121
• Datasets 3.0.2
• Tokenizers 0.19.1