ImageNette resnet50-pytorch

Model Description

resnet50-pytorch trained on 10-class ImageNet subset (ImageNette) with advanced augmentation techniques.

Model Architecture

Architecture: resnet50-pytorch
Dataset: ImageNette
Classes: 10

Training Configuration

Batch Size: 128
Optimizer: sgd (momentum=0.9, weight_decay=1e-3)
Scheduler: onecycle
Augmentation: HorizontalFlip, ShiftScaleRotate, Cutout, ColorJitter
MixUp: Alpha=0.2
Label Smoothing: 0.1
Mixed Precision: True
Gradient Clipping: 1.0

Performance

Best Test Accuracy: 57.30%
Total Epochs Trained: 10
Final Train Accuracy: 46.22%
Final Test Accuracy: 57.30%

Training History

Best Epoch: 10
Train Loss: 2.4472 → 1.7581
Test Loss: 3.4594 → 1.3567

Usage

import torch
from huggingface_hub import hf_hub_download

# Download model
checkpoint_path = hf_hub_download(
    repo_id="pandurangpatil/imagenet10trial",
    filename="best_model.pth"
)

# Load checkpoint
checkpoint = torch.load(checkpoint_path, map_location='cpu', weights_only=False)

# Load model (you'll need to have the model definition)
# from models import get_model
# model = get_model('resnet50-pytorch', num_classes=10)
# model.load_state_dict(checkpoint['model_state_dict'])
# model.eval()

Training Details

Dataset: ImageNette (9469 train, 3925 test)
Classes: 10
Image Size: 160×160 or 224×224
Normalization: mean=(0.485, 0.456, 0.406), std=(0.229, 0.224, 0.225)

Files

best_model.pth - Best performing model checkpoint
training_curves.png - Training/test accuracy and loss curves
lr_finder_plot.png - Learning rate finder results
metrics.json - Complete training history
config.json - Hyperparameter configuration

License

MIT

Citation

@misc{resnet50-pytorch-imagenette,
  title = {ImageNette resnet50-pytorch},
  year = {2025},
  publisher = {HuggingFace},
  url = {https://huggingface.co/pandurangpatil/imagenet10trial}
}

Downloads last month: 10