test_triage

This model is a fine-tuned version of microsoft/swin-tiny-patch4-window7-224 on the Test dataset. It achieves the following results on the evaluation set:

• Loss: 1.9758
• Accuracy: 0.5008

Model description

This is a basic skin disease recognition model without the specific disease information for now. I just wanted to test the platform for hosting capabilities and check other features.

Intended uses & limitations

For now, its just a test environment. We have the basic pipeline of data & processing in place to push to this place. Future use will be to open source the dataset and allow the community to fine tune the skin identification and triaging module for broader and free-for-all in commercial/non-commercial usage.

Training and evaluation data

We have a lot of open & closed datasets that have been compiled over years and annotated.

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 5e-05
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 128
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 6

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
3.3471	1.0	151	3.2152	0.2452
2.7313	2.0	303	2.5291	0.3817
2.48	3.0	454	2.2459	0.4413
2.2192	4.0	606	2.0968	0.4702
2.0479	5.0	757	2.0026	0.4897
1.9702	5.98	906	1.9758	0.5008

Framework versions

• Transformers 4.30.2
• Pytorch 2.0.1+cu118
• Datasets 2.13.1
• Tokenizers 0.13.3