[2021-03-20 00:28:02] Created experiment 0:
[2021-03-20 00:28:02]  - Model: vgg-pretrained
[2021-03-20 00:28:02]  - Acquisition function: lc-beam-coreset
[2021-03-20 00:28:02] Created experiment 1:
[2021-03-20 00:28:02]  - Model: vgg-pretrained
[2021-03-20 00:28:02]  - Acquisition function: greedy-coreset
[2021-03-20 00:28:02] Created experiment 2:
[2021-03-20 00:28:02]  - Model: vgg-pretrained
[2021-03-20 00:28:02]  - Acquisition function: random
[2021-03-20 00:28:02] Loading cifar10 test set...
[2021-03-20 00:28:03] Experiment repeat 1/1
[2021-03-20 00:28:03] Seeded: 5
[2021-03-20 00:28:03] Using 2.00% labels of the dataset (1000/50000)
[2021-03-20 00:28:03] Randomly labelled 1000/50000
[2021-03-20 00:28:03] Showing labelled: True (1000/50000 visible, 0 redundant)
[2021-03-20 00:28:03] Seeded: 6
[2021-03-20 00:28:03] Running: experiment 0
[2021-03-20 00:28:03] Showing labelled: True (1000/50000 visible, 0 redundant)
[2021-03-20 00:28:03] Creating pretrained=True VGG16...
[2021-03-20 00:28:05] No parameter reset since we are using a pretrained model.
[2021-03-20 00:28:05] vgg-pretrained: initialized 15245130 parameters.
[2021-03-20 00:28:05] Creating trainer with model on device: cuda
[2021-03-20 00:28:10] Training vgg-pretrained across 1000 data points in cifar10...
[2021-03-20 00:28:32] Training accuracy: 0.9874
[2021-03-20 00:28:32] Testing on 10000 data points...
[2021-03-20 00:28:34] Test score for 1000 training labels: 0.7204
[2021-03-20 00:28:34] Showing labelled: False (50000/50000 visible, 0 redundant)
[2021-03-20 00:28:34] Found 50000 unlabelled features.
[2021-03-20 00:28:45] Computing distance between 1000 labelled and 50000 unlabelled vectors of length 512...
[2021-03-20 00:28:48] Searching for coresets with 20 beams...
[2021-03-20 00:29:16] Showing labelled: True (1400/50000 visible, 0 redundant)
[2021-03-20 00:29:16] Creating trainer with model on device: cuda
[2021-03-20 00:29:16] Training vgg-pretrained across 1400 data points in cifar10...
[2021-03-20 00:29:24] Training accuracy: 0.9745
[2021-03-20 00:29:24] Testing on 10000 data points...
[2021-03-20 00:29:26] Test score for 1400 training labels: 0.7405
[2021-03-20 00:29:26] Showing labelled: False (50000/50000 visible, 0 redundant)
[2021-03-20 00:29:26] Found 50000 unlabelled features.
[2021-03-20 00:29:37] Computing distance between 1400 labelled and 50000 unlabelled vectors of length 512...
[2021-03-20 00:29:41] Searching for coresets with 20 beams...
[2021-03-20 00:30:09] Showing labelled: True (1800/50000 visible, 0 redundant)
[2021-03-20 00:30:09] Creating trainer with model on device: cuda
[2021-03-20 00:30:09] Training vgg-pretrained across 1800 data points in cifar10...
[2021-03-20 00:30:20] Training accuracy: 0.9797
[2021-03-20 00:30:20] Testing on 10000 data points...
[2021-03-20 00:30:22] Test score for 1800 training labels: 0.7607
[2021-03-20 00:30:22] Showing labelled: False (50000/50000 visible, 0 redundant)
[2021-03-20 00:30:22] Found 50000 unlabelled features.
[2021-03-20 00:30:33] Computing distance between 1800 labelled and 50000 unlabelled vectors of length 512...
[2021-03-20 00:30:38] Searching for coresets with 20 beams...
[2021-03-20 00:31:06] Showing labelled: True (2200/50000 visible, 0 redundant)
[2021-03-20 00:31:06] Creating trainer with model on device: cuda
[2021-03-20 00:31:06] Training vgg-pretrained across 2200 data points in cifar10...
[2021-03-20 00:31:20] Training accuracy: 0.9891
[2021-03-20 00:31:20] Testing on 10000 data points...
[2021-03-20 00:31:22] Test score for 2200 training labels: 0.7773
[2021-03-20 00:31:22] Showing labelled: False (50000/50000 visible, 0 redundant)
[2021-03-20 00:31:22] Found 50000 unlabelled features.
[2021-03-20 00:31:33] Computing distance between 2200 labelled and 50000 unlabelled vectors of length 512...
[2021-03-20 00:31:39] Searching for coresets with 20 beams...
[2021-03-20 00:32:08] Showing labelled: True (2600/50000 visible, 0 redundant)
[2021-03-20 00:32:08] Creating trainer with model on device: cuda
[2021-03-20 00:32:08] Training vgg-pretrained across 2600 data points in cifar10...
[2021-03-20 00:32:23] Training accuracy: 0.9917
[2021-03-20 00:32:23] Testing on 10000 data points...
[2021-03-20 00:32:25] Test score for 2600 training labels: 0.7863
[2021-03-20 00:32:25] Showing labelled: False (50000/50000 visible, 0 redundant)
[2021-03-20 00:32:25] Found 50000 unlabelled features.
[2021-03-20 00:32:36] Computing distance between 2600 labelled and 50000 unlabelled vectors of length 512...
[2021-03-20 00:32:44] Searching for coresets with 20 beams...
[2021-03-20 00:33:12] Showing labelled: True (3000/50000 visible, 0 redundant)
[2021-03-20 00:33:12] Creating trainer with model on device: cuda
[2021-03-20 00:33:12] Training vgg-pretrained across 3000 data points in cifar10...
[2021-03-20 00:33:31] Training accuracy: 0.9934
[2021-03-20 00:33:31] Testing on 10000 data points...
[2021-03-20 00:33:33] Test score for 3000 training labels: 0.7923
[2021-03-20 00:33:33] Showing labelled: False (50000/50000 visible, 0 redundant)
[2021-03-20 00:33:33] Found 50000 unlabelled features.
[2021-03-20 00:33:44] Computing distance between 3000 labelled and 50000 unlabelled vectors of length 512...
[2021-03-20 00:33:53] Searching for coresets with 20 beams...
[2021-03-20 00:34:21] Showing labelled: True (3400/50000 visible, 0 redundant)
[2021-03-20 00:34:21] Creating trainer with model on device: cuda
[2021-03-20 00:34:21] Training vgg-pretrained across 3400 data points in cifar10...
[2021-03-20 00:34:42] Training accuracy: 0.9946
[2021-03-20 00:34:42] Testing on 10000 data points...
[2021-03-20 00:34:44] Test score for 3400 training labels: 0.8032
[2021-03-20 00:34:44] Showing labelled: False (50000/50000 visible, 0 redundant)
[2021-03-20 00:34:44] Found 50000 unlabelled features.
[2021-03-20 00:34:55] Computing distance between 3400 labelled and 50000 unlabelled vectors of length 512...
[2021-03-20 00:35:05] Searching for coresets with 20 beams...
[2021-03-20 00:35:33] Showing labelled: True (3800/50000 visible, 0 redundant)
[2021-03-20 00:35:33] Creating trainer with model on device: cuda
[2021-03-20 00:35:33] Training vgg-pretrained across 3800 data points in cifar10...
[2021-03-20 00:35:57] Training accuracy: 0.9947
[2021-03-20 00:35:57] Testing on 10000 data points...
[2021-03-20 00:35:59] Test score for 3800 training labels: 0.8075
[2021-03-20 00:35:59] Showing labelled: False (50000/50000 visible, 0 redundant)
[2021-03-20 00:35:59] Found 50000 unlabelled features.
[2021-03-20 00:36:11] Computing distance between 3800 labelled and 50000 unlabelled vectors of length 512...
[2021-03-20 00:36:21] Searching for coresets with 20 beams...
[2021-03-20 00:36:50] Showing labelled: True (4200/50000 visible, 0 redundant)
[2021-03-20 00:36:50] Creating trainer with model on device: cuda
[2021-03-20 00:36:50] Training vgg-pretrained across 4200 data points in cifar10...
[2021-03-20 00:37:15] Training accuracy: 0.9983
[2021-03-20 00:37:15] Testing on 10000 data points...
[2021-03-20 00:37:17] Test score for 4200 training labels: 0.8102
[2021-03-20 00:37:17] Showing labelled: False (50000/50000 visible, 0 redundant)
[2021-03-20 00:37:17] Found 50000 unlabelled features.
[2021-03-20 00:37:29] Computing distance between 4200 labelled and 50000 unlabelled vectors of length 512...
[2021-03-20 00:37:41] Searching for coresets with 20 beams...
[2021-03-20 00:38:09] Showing labelled: True (4600/50000 visible, 0 redundant)
[2021-03-20 00:38:09] Creating trainer with model on device: cuda
[2021-03-20 00:38:09] Training vgg-pretrained across 4600 data points in cifar10...
[2021-03-20 00:38:37] Training accuracy: 0.9988
[2021-03-20 00:38:37] Testing on 10000 data points...
[2021-03-20 00:38:39] Test score for 4600 training labels: 0.8210
[2021-03-20 00:38:39] Showing labelled: False (50000/50000 visible, 0 redundant)
[2021-03-20 00:38:39] Found 50000 unlabelled features.
[2021-03-20 00:38:51] Computing distance between 4600 labelled and 50000 unlabelled vectors of length 512...
[2021-03-20 00:39:05] Searching for coresets with 20 beams...
[2021-03-20 00:39:33] Showing labelled: True (5000/50000 visible, 0 redundant)
[2021-03-20 00:39:33] Creating trainer with model on device: cuda
[2021-03-20 00:39:33] Training vgg-pretrained across 5000 data points in cifar10...
[2021-03-20 00:40:03] Training accuracy: 0.9984
[2021-03-20 00:40:03] Testing on 10000 data points...
[2021-03-20 00:40:05] Test score for 5000 training labels: 0.8230
[2021-03-20 00:40:05] Seeded: 6
[2021-03-20 00:40:05] Running: experiment 1
[2021-03-20 00:40:05] Showing labelled: True (1000/50000 visible, 0 redundant)
[2021-03-20 00:40:05] Creating pretrained=True VGG16...
[2021-03-20 00:40:07] No parameter reset since we are using a pretrained model.
[2021-03-20 00:40:07] vgg-pretrained: initialized 15245130 parameters.
[2021-03-20 00:40:07] Creating trainer with model on device: cuda
[2021-03-20 00:40:07] Training vgg-pretrained across 1000 data points in cifar10...
[2021-03-20 00:40:07] Updated results: ../results/cifar10/vgg_pretrained_gcoreset/results.json
