cvmil/vit-hybrid-base-bit-384_augmented-v2_tl

vit-hybrid-base-bit-384_rice-leaf-disease-augmented-v2_tl

This model is a fine-tuned version of google/vit-hybrid-base-bit-384 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.5725
• Accuracy: 0.8185

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: 0.0003
• train_batch_size: 128
• eval_batch_size: 128
• seed: 42
• 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: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
2.0186	1.0	63	1.7057	0.4256
1.3213	2.0	126	1.0837	0.6756
0.8384	3.0	189	0.8425	0.7411
0.6553	4.0	252	0.7551	0.7738
0.5601	5.0	315	0.6954	0.7768
0.5006	6.0	378	0.6683	0.7798
0.4592	7.0	441	0.6419	0.7976
0.4272	8.0	504	0.6246	0.8095
0.4029	9.0	567	0.6096	0.8036
0.3844	10.0	630	0.6018	0.7976
0.3685	11.0	693	0.5945	0.8125
0.3566	12.0	756	0.5869	0.8095
0.3486	13.0	819	0.5806	0.8155
0.3405	14.0	882	0.5764	0.8214
0.3351	15.0	945	0.5758	0.8185
0.3314	16.0	1008	0.5751	0.8185
0.3296	17.0	1071	0.5738	0.8185
0.327	18.0	1134	0.5726	0.8185
0.3265	19.0	1197	0.5726	0.8185
0.326	20.0	1260	0.5725	0.8185

Framework versions

• Transformers 4.48.3
• Pytorch 2.5.1+cu124
• Datasets 3.3.2
• Tokenizers 0.21.0