hiera-finetuned-stroke-multi

This model is a fine-tuned version of BTX24/hiera-base-224-in1k-hf-finetuned-stroke-binary on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.0648
• Accuracy: 0.9861
• F1: 0.9861
• Precision: 0.9862
• Recall: 0.9861

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: 2e-05
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 64
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine_with_restarts
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 12
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1	Precision	Recall
0.0612	1.0997	100	0.1259	0.9639	0.9635	0.9647	0.9639
0.0694	2.1994	200	0.0735	0.9778	0.9777	0.9780	0.9778
0.0632	3.2992	300	0.0867	0.9764	0.9763	0.9765	0.9764
0.064	4.3989	400	0.0677	0.9827	0.9826	0.9826	0.9827
0.0614	5.4986	500	0.0766	0.9827	0.9827	0.9830	0.9827
0.0361	6.5983	600	0.1001	0.9764	0.9763	0.9764	0.9764
0.0321	7.6981	700	0.0701	0.9827	0.9827	0.9828	0.9827
0.0495	8.7978	800	0.0648	0.9861	0.9861	0.9862	0.9861
0.0339	9.8975	900	0.0615	0.9854	0.9854	0.9855	0.9854
0.0384	10.9972	1000	0.0628	0.9854	0.9854	0.9855	0.9854

Framework versions

• Transformers 4.52.4
• Pytorch 2.7.1+cu126
• Datasets 3.6.0
• Tokenizers 0.21.1