Views
No views yet
| Model | Macro-F1 | Params (M) | Fold |
|---|---|---|---|
| ShuffleNet V2 | 0.648 | 1.26 | 3 |
| DenseNet-121 | 0.637 | 6.96 | 0 |
| ViT-Small | 0.624 | 21.67 | 4 |
| SE-ResNet50 | 0.617 | 26.05 | 1 |
| EfficientNetV2-S | 0.615 | 20.19 | 0 |
| ResNet-50 | 0.604 | 23.52 | 3 |
| RegNetY-032 | 0.608 | 17.93 | 4 |
| MobileNetV3 | 0.631 | 4.21 | 1 |
| ConvNeXt Tiny | 0.601 | 27.82 | 0 |
| EfficientNet-B0 | 0.613 | 4.02 | 0 |
1from huggingface_hub import hf_hub_download
2import torch
3
4# Download a specific model
5ckpt_path = hf_hub_download(
6 repo_id="muhwira27/skin-lesion-models",
7 filename="shufflenet_v2_x1_0_best.ckpt"
8)
9
10# Load checkpoint
11checkpoint = torch.load(ckpt_path, map_location="cpu")
12
13# Checkpoint contents
14# - backbone: str (model architecture name)
15# - state_dict: model weights
16# - label2id: dict mapping class names to indices
17# - img_size: int (input image size, typically 224)shufflenet_v2_x1_0_best.ckpt - Best model (smallest & best performance)densenet121_best.ckptvit_small_patch16_224_best.ckptseresnet50_best.ckpttf_efficientnetv2_s_best.ckptresnet50_best.ckptregnety_032_best.ckptmobilenet_v3_large_best.ckptconvnext_tiny_best.ckptefficientnet_b0_best.ckpt