Models Directory

Place your trained model checkpoints here.

Usage

Copy from Training Directory

# Copy a trained checkpoint
cp /home/ubuntu/imagenet/checkpoints/resnet50-epoch=89-val_acc1=0.7500.ckpt .

# Or copy the latest checkpoint
cp /home/ubuntu/imagenet/checkpoints/last.ckpt .

Use in Docker

When running the Streamlit app in Docker, reference the checkpoint as:

/app/models/resnet50-epoch=89-val_acc1=0.7500.ckpt

The models/ directory is mounted as a volume in Docker, so you can:

Add/update checkpoints without rebuilding the container
Access them from the Streamlit UI

Supported Formats

PyTorch Lightning checkpoints (.ckpt)
Standard PyTorch checkpoints (.pth, .pt)

Example

# List available checkpoints
ls -lh

# Test inference with a checkpoint
cd ..
python inference.py \
    --image test_image.jpg \
    --checkpoint models/resnet50-epoch=89.ckpt \
    --verbose