Skip to content

Latest commit

 

History

History
44 lines (28 loc) · 1.02 KB

README.md

File metadata and controls

44 lines (28 loc) · 1.02 KB
Raw

Deep Multiple Instance Learning with Distance-Aware Self-Attention (DAS-MIL)

Paper link

Installation

If you're using a GPU, run:

./install_gpu.sh

Otherwise, use:

./install_cpu.sh

Training

MNIST-COLLAGE

python3 train.py +selected_model/mnist_collage=distance_aware_self_attention +experiment=mnist_collage

MNIST-COLLAGE-INV

python3 train.py +selected_model/mnist_collage=distance_aware_self_attention +experiment=mnist_collage_inverse

CAMELYON16

python3 train.py +selected_model/camelyon16=distance_aware_self_attention

You can use the option device=0 to use GPU 0. Full list of options is available at conf/config.yaml. We use hydra for configuration management.

Ablations

See the YAML files in conf/selected_model/mnist_collage_ablations. Run ./run_trials_for_selected_models_cpu.sh to run all the ablations (set selected_model_type="mnist_collage_ablations" in the shell script).