Dacon 악성 URL 분류 AI 경진대회

링크

image/png

base_model : distilbert/distilbert-base-uncased

tune parameters

batch_size : 64
epocs : 5
learning_rate : 2e-5
weight_decay : 0.01

Model Training Results

Step Training Loss Validation Loss Accuracy F1 Score
200 0.024300 0.453651 0.929750 0.928860
400 0.016700 0.543543 0.917750 0.916868
600 0.007600 0.604909 0.921250 0.921414
800 0.008500 0.588405 0.922750 0.922533
1000 0.007300 0.618596 0.925500 0.925138
1200 0.006300 0.628956 0.923250 0.922644

PS: 성능치 좋으나 아직 테스트를 하지 않은 상태라 추구 인퍼런스 필요

Downloads last month
17
Safetensors
Model size
67M params
Tensor type
F32
·
Inference Providers NEW
This model is not currently available via any of the supported Inference Providers.

Model tree for UICHEOL-HWANG/malicious-url-detection

Finetuned
(7498)
this model