search

MedMNIST / experiments

Codebase for reproducible benchmarking experiments in MedMNIST v2
Apache License 2.0
44 stars 18 forks source link

Training Settings for MedMNIST Experiments #12

Open KateElizabeth112 opened 1 week ago

KateElizabeth112 commented 1 week ago

Hello,

I would like to reproduce the experiments that led to your ResNet results in the table below. Are you able to share the command line argument settings that you used to train and evaluate each model, for example, --num_epochs ?

Many thanks, Kate

Screenshot 2024-10-23 at 11 07 13

duducheng commented 1 week ago

Hi Kate ,

I think they should be reproduced by default. Can you confirm this? @ThreeSRR

JC

ThreeSRR commented 1 week ago

The results can be reproduced by default.

Just like python train_and_eval_pytorch.py --data_flag pathmnist for three-channel datasets, or python train_and_eval_pytorch.py --data_flag breastmnist --as_rgb for single-channel ones.