imagenet dataset issue

[farooq199]

For the mini_imagenet dataset, after the images are inflated the program terminates automatically when counter reaches 60000 count

[vgsatorras]

Hi,

What mini_imagenet file are you using to run it? If you send an email to [e-mail] I can send you the compressed mini_imagenet dataset that runs with no problems.

If you already did that and the problem persists, could you copy the end of the log file to better assess what is the problem? The code runs without crashing in my machine.

Best,
Victor

[farooq199]

The issue was with the size of the dataset, which is now resolved. Also, can you provide the experiment name for the active learning scenario and the values of the parameters used, as we have to reproduce your results

[vgsatorras]

Hi,

The commands to run the active learning experiments are the following

# Active learning
EXPNAME=minimagenet_N5_S1_U4_active
python3 -u main.py --exp_name $EXPNAME --dataset mini_imagenet --test_N_way 5 --train_N_way 5 --train_N_shots 5 --test_N_shots 5 --unlabeled_extra 4 --metric_network gnn_iclr_active --batch_size 40 --active_random 0 --dec_lr=15000 --iterations 90000 --test_interval 2000

# Random baseline in the Active learning experiment
EXPNAME=minimagenet_N5_S1_U4_active_random
python3 -u main.py --exp_name $EXPNAME --dataset mini_imagenet --test_N_way 5 --train_N_way 5 --train_N_shots 5 --test_N_shots 5 --unlabeled_extra 4 --metric_network gnn_iclr_active --active_random 1 --batch_size 40 --dec_lr=15000 --iterations 90000 --test_interval 2000

You may have to pull the repository again in order to run them since I just uploaded that part of the code.

Best,
Victor