material-surface-classifier

This model is a fine-tuned version of timm/mobilenetv3_large_100.ra_in1k on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.5208
• Accuracy: 0.83
• F1 Macro: 0.7133

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: 16
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 64
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 0.1
• num_epochs: 15

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1 Macro
1.3096	1.0	25	1.1979	0.54	0.4446
0.7334	2.0	50	0.6755	0.7075	0.6149
0.6720	3.0	75	0.5615	0.75	0.6328
0.5520	4.0	100	0.4911	0.7875	0.6849
0.5370	5.0	125	0.4791	0.7875	0.6668
0.4934	6.0	150	0.4929	0.825	0.7121
0.4253	7.0	175	0.4966	0.8325	0.7120
0.3215	8.0	200	0.4997	0.8175	0.7296
0.3122	9.0	225	0.4815	0.835	0.7263
0.2824	10.0	250	0.4749	0.83	0.7124
0.2727	11.0	275	0.5188	0.835	0.7255
0.1778	12.0	300	0.4973	0.8225	0.7058
0.2922	13.0	325	0.4867	0.8425	0.7355
0.2612	14.0	350	0.5655	0.85	0.7334
0.2593	15.0	375	0.5208	0.83	0.7133

Framework versions

• Transformers 5.7.0
• Pytorch 2.11.0+cu130
• Datasets 4.8.5
• Tokenizers 0.22.2