[2021-03-22 20:11:32] Created experiment 0:
[2021-03-22 20:11:32]  - Model: vgg-pt
[2021-03-22 20:11:32]  - Acquisition function: lc-beam-pweighted-coreset (beams=20)
[2021-03-22 20:11:32] Loading cifar100 test set...
[2021-03-22 20:11:33] Experiment repeat 1/5
[2021-03-22 20:11:33] Seeded: 5
[2021-03-22 20:11:33] Using 10.00% labels of the dataset (5000/50000)
[2021-03-22 20:11:33] Randomly labelled 5000/50000
[2021-03-22 20:11:33] Showing labelled: True (5000/50000 visible, 0 redundant)
[2021-03-22 20:11:33] Seeded: 6
[2021-03-22 20:11:33] Running: experiment 0
[2021-03-22 20:11:33] Showing labelled: True (5000/50000 visible, 0 redundant)
[2021-03-22 20:11:33] Creating pretrained=True VGG16...
[2021-03-22 20:11:35] No parameter reset since we are using a pretrained model.
[2021-03-22 20:11:35] vgg-pretrained: initialized 15291300 parameters.
[2021-03-22 20:11:35] Creating trainer with model on device: cuda
[2021-03-22 20:11:40] Training with expected score >= 0.9900
[2021-03-22 20:11:40] Training vgg-pretrained across 5000 data points in cifar100...
[2021-03-22 20:16:43] Training accuracy: 0.9973
[2021-03-22 20:16:43] Testing on 10000 data points...
[2021-03-22 20:16:45] Test score for 5000 training labels: 0.4888
[2021-03-22 20:16:45] Showing labelled: False (50000/50000 visible, 0 redundant)
[2021-03-22 20:16:45] Found 50000 unlabelled features.
[2021-03-22 20:16:57] Computing distance between 5000 labelled and 50000 unlabelled vectors of length 512...
[2021-03-22 20:17:12] Searching for coresets with 20 beams...
[2021-03-22 20:22:38] Showing labelled: True (10000/50000 visible, 0 redundant)
[2021-03-22 20:22:38] Creating trainer with model on device: cuda
[2021-03-22 20:22:38] Training with expected score >= 0.9900
[2021-03-22 20:22:38] Training vgg-pretrained across 10000 data points in cifar100...
[2021-03-22 20:25:46] Training accuracy: 0.9939
[2021-03-22 20:25:46] Testing on 10000 data points...
[2021-03-22 20:25:48] Test score for 10000 training labels: 0.5483
[2021-03-22 20:25:48] Showing labelled: False (50000/50000 visible, 0 redundant)
[2021-03-22 20:25:48] Found 50000 unlabelled features.
[2021-03-22 20:25:59] Computing distance between 10000 labelled and 50000 unlabelled vectors of length 512...
[2021-03-22 20:26:28] Searching for coresets with 20 beams...
[2021-03-22 20:31:52] Showing labelled: True (15000/50000 visible, 0 redundant)
[2021-03-22 20:31:52] Creating trainer with model on device: cuda
[2021-03-22 20:31:52] Training with expected score >= 0.9900
[2021-03-22 20:31:52] Training vgg-pretrained across 15000 data points in cifar100...
[2021-03-22 20:36:35] Training accuracy: 0.9980
[2021-03-22 20:36:35] Testing on 10000 data points...
[2021-03-22 20:36:37] Test score for 15000 training labels: 0.5825
[2021-03-22 20:36:37] Showing labelled: False (50000/50000 visible, 0 redundant)
[2021-03-22 20:36:37] Found 50000 unlabelled features.
[2021-03-22 20:36:50] Computing distance between 15000 labelled and 50000 unlabelled vectors of length 512...
[2021-03-22 20:37:33] Searching for coresets with 20 beams...
[2021-03-22 20:42:57] Showing labelled: True (20000/50000 visible, 0 redundant)
[2021-03-22 20:42:57] Creating trainer with model on device: cuda
[2021-03-22 20:42:57] Training with expected score >= 0.9900
[2021-03-22 20:42:57] Training vgg-pretrained across 20000 data points in cifar100...
[2021-03-22 20:49:12] Training accuracy: 0.9981
[2021-03-22 20:49:12] Testing on 10000 data points...
[2021-03-22 20:49:15] Test score for 20000 training labels: 0.5978
[2021-03-22 20:49:15] Experiment repeat 2/5
[2021-03-22 20:49:15] Seeded: 6
[2021-03-22 20:49:15] Using 10.00% labels of the dataset (5000/50000)
[2021-03-22 20:49:15] Randomly labelled 5000/50000
[2021-03-22 20:49:15] Showing labelled: True (5000/50000 visible, 0 redundant)
[2021-03-22 20:49:15] Seeded: 7
[2021-03-22 20:49:15] Running: experiment 0
[2021-03-22 20:49:15] Showing labelled: True (5000/50000 visible, 0 redundant)
[2021-03-22 20:49:15] Creating pretrained=True VGG16...
[2021-03-22 20:49:16] No parameter reset since we are using a pretrained model.
[2021-03-22 20:49:16] vgg-pretrained: initialized 15291300 parameters.
[2021-03-22 20:49:16] Creating trainer with model on device: cuda
[2021-03-22 20:49:16] Training with expected score >= 0.9900
[2021-03-22 20:49:16] Training vgg-pretrained across 5000 data points in cifar100...
[2021-03-22 20:54:19] Training accuracy: 0.9975
[2021-03-22 20:54:19] Testing on 10000 data points...
[2021-03-22 20:54:21] Test score for 5000 training labels: 0.4832
[2021-03-22 20:54:21] Showing labelled: False (50000/50000 visible, 0 redundant)
[2021-03-22 20:54:21] Found 50000 unlabelled features.
[2021-03-22 20:54:33] Computing distance between 5000 labelled and 50000 unlabelled vectors of length 512...
[2021-03-22 20:54:47] Searching for coresets with 20 beams...
[2021-03-22 21:00:12] Showing labelled: True (10000/50000 visible, 0 redundant)
[2021-03-22 21:00:12] Creating trainer with model on device: cuda
[2021-03-22 21:00:12] Training with expected score >= 0.9900
[2021-03-22 21:00:12] Training vgg-pretrained across 10000 data points in cifar100...
[2021-03-22 21:03:21] Training accuracy: 0.9950
[2021-03-22 21:03:21] Testing on 10000 data points...
[2021-03-22 21:03:23] Test score for 10000 training labels: 0.5326
[2021-03-22 21:03:23] Showing labelled: False (50000/50000 visible, 0 redundant)
[2021-03-22 21:03:23] Found 50000 unlabelled features.
[2021-03-22 21:03:35] Computing distance between 10000 labelled and 50000 unlabelled vectors of length 512...
[2021-03-22 21:04:06] Searching for coresets with 20 beams...
[2021-03-22 21:09:31] Showing labelled: True (15000/50000 visible, 0 redundant)
[2021-03-22 21:09:31] Creating trainer with model on device: cuda
[2021-03-22 21:09:31] Training with expected score >= 0.9900
[2021-03-22 21:09:31] Training vgg-pretrained across 15000 data points in cifar100...
[2021-03-22 21:14:12] Training accuracy: 0.9993
[2021-03-22 21:14:12] Testing on 10000 data points...
[2021-03-22 21:14:14] Test score for 15000 training labels: 0.5750
[2021-03-22 21:14:14] Showing labelled: False (50000/50000 visible, 0 redundant)
[2021-03-22 21:14:14] Found 50000 unlabelled features.
[2021-03-22 21:14:28] Computing distance between 15000 labelled and 50000 unlabelled vectors of length 512...
[2021-03-22 21:15:15] Searching for coresets with 20 beams...
[2021-03-22 21:20:40] Showing labelled: True (20000/50000 visible, 0 redundant)
[2021-03-22 21:20:40] Creating trainer with model on device: cuda
[2021-03-22 21:20:40] Training with expected score >= 0.9900
[2021-03-22 21:20:40] Training vgg-pretrained across 20000 data points in cifar100...
[2021-03-22 21:26:55] Training accuracy: 0.9981
[2021-03-22 21:26:55] Testing on 10000 data points...
[2021-03-22 21:26:57] Test score for 20000 training labels: 0.5940
[2021-03-22 21:26:57] Experiment repeat 3/5
[2021-03-22 21:26:57] Seeded: 7
[2021-03-22 21:26:57] Using 10.00% labels of the dataset (5000/50000)
[2021-03-22 21:26:57] Randomly labelled 5000/50000
[2021-03-22 21:26:57] Showing labelled: True (5000/50000 visible, 0 redundant)
[2021-03-22 21:26:57] Seeded: 8
[2021-03-22 21:26:57] Running: experiment 0
[2021-03-22 21:26:57] Showing labelled: True (5000/50000 visible, 0 redundant)
[2021-03-22 21:26:57] Creating pretrained=True VGG16...
[2021-03-22 21:26:58] No parameter reset since we are using a pretrained model.
[2021-03-22 21:26:58] vgg-pretrained: initialized 15291300 parameters.
[2021-03-22 21:26:58] Creating trainer with model on device: cuda
[2021-03-22 21:26:58] Training with expected score >= 0.9900
[2021-03-22 21:26:58] Training vgg-pretrained across 5000 data points in cifar100...
[2021-03-22 21:32:02] Training accuracy: 0.9987
[2021-03-22 21:32:02] Testing on 10000 data points...
[2021-03-22 21:32:04] Test score for 5000 training labels: 0.4881
[2021-03-22 21:32:04] Showing labelled: False (50000/50000 visible, 0 redundant)
[2021-03-22 21:32:04] Found 50000 unlabelled features.
[2021-03-22 21:32:16] Computing distance between 5000 labelled and 50000 unlabelled vectors of length 512...
[2021-03-22 21:32:30] Searching for coresets with 20 beams...
[2021-03-22 21:37:55] Showing labelled: True (10000/50000 visible, 0 redundant)
[2021-03-22 21:37:55] Creating trainer with model on device: cuda
[2021-03-22 21:37:55] Training with expected score >= 0.9900
[2021-03-22 21:37:55] Training vgg-pretrained across 10000 data points in cifar100...
[2021-03-22 21:41:04] Training accuracy: 0.9900
[2021-03-22 21:41:04] Testing on 10000 data points...
[2021-03-22 21:41:06] Test score for 10000 training labels: 0.5444
[2021-03-22 21:41:06] Showing labelled: False (50000/50000 visible, 0 redundant)
[2021-03-22 21:41:06] Found 50000 unlabelled features.
[2021-03-22 21:41:19] Computing distance between 10000 labelled and 50000 unlabelled vectors of length 512...
[2021-03-22 21:41:50] Searching for coresets with 20 beams...
[2021-03-22 21:47:15] Showing labelled: True (15000/50000 visible, 0 redundant)
[2021-03-22 21:47:15] Creating trainer with model on device: cuda
[2021-03-22 21:47:15] Training with expected score >= 0.9900
[2021-03-22 21:47:15] Training vgg-pretrained across 15000 data points in cifar100...
[2021-03-22 21:51:56] Training accuracy: 0.9960
[2021-03-22 21:51:56] Testing on 10000 data points...
[2021-03-22 21:51:58] Test score for 15000 training labels: 0.5777
[2021-03-22 21:51:58] Showing labelled: False (50000/50000 visible, 0 redundant)
[2021-03-22 21:51:58] Found 50000 unlabelled features.
[2021-03-22 21:52:11] Computing distance between 15000 labelled and 50000 unlabelled vectors of length 512...
[2021-03-22 21:52:57] Searching for coresets with 20 beams...
[2021-03-22 21:58:22] Showing labelled: True (20000/50000 visible, 0 redundant)
[2021-03-22 21:58:22] Creating trainer with model on device: cuda
[2021-03-22 21:58:22] Training with expected score >= 0.9900
[2021-03-22 21:58:22] Training vgg-pretrained across 20000 data points in cifar100...
[2021-03-22 22:04:37] Training accuracy: 0.9986
[2021-03-22 22:04:37] Testing on 10000 data points...
[2021-03-22 22:04:39] Test score for 20000 training labels: 0.5964
[2021-03-22 22:04:39] Experiment repeat 4/5
[2021-03-22 22:04:39] Seeded: 8
[2021-03-22 22:04:39] Using 10.00% labels of the dataset (5000/50000)
[2021-03-22 22:04:39] Randomly labelled 5000/50000
[2021-03-22 22:04:39] Showing labelled: True (5000/50000 visible, 0 redundant)
[2021-03-22 22:04:39] Seeded: 9
[2021-03-22 22:04:39] Running: experiment 0
[2021-03-22 22:04:39] Showing labelled: True (5000/50000 visible, 0 redundant)
[2021-03-22 22:04:39] Creating pretrained=True VGG16...
[2021-03-22 22:04:41] No parameter reset since we are using a pretrained model.
[2021-03-22 22:04:41] vgg-pretrained: initialized 15291300 parameters.
[2021-03-22 22:04:41] Creating trainer with model on device: cuda
[2021-03-22 22:04:41] Training with expected score >= 0.9900
[2021-03-22 22:04:41] Training vgg-pretrained across 5000 data points in cifar100...
[2021-03-22 22:09:44] Training accuracy: 0.9974
[2021-03-22 22:09:44] Testing on 10000 data points...
[2021-03-22 22:09:46] Test score for 5000 training labels: 0.4916
[2021-03-22 22:09:46] Showing labelled: False (50000/50000 visible, 0 redundant)
[2021-03-22 22:09:46] Found 50000 unlabelled features.
[2021-03-22 22:09:58] Computing distance between 5000 labelled and 50000 unlabelled vectors of length 512...
[2021-03-22 22:10:12] Searching for coresets with 20 beams...
[2021-03-22 22:17:48] Showing labelled: True (10000/50000 visible, 0 redundant)
[2021-03-22 22:17:48] Creating trainer with model on device: cuda
[2021-03-22 22:17:48] Training with expected score >= 0.9900
[2021-03-22 22:17:48] Training vgg-pretrained across 10000 data points in cifar100...
[2021-03-22 22:20:55] Training accuracy: 0.9929
[2021-03-22 22:20:55] Testing on 10000 data points...
[2021-03-22 22:20:57] Test score for 10000 training labels: 0.5455
[2021-03-22 22:20:57] Showing labelled: False (50000/50000 visible, 0 redundant)
[2021-03-22 22:20:57] Found 50000 unlabelled features.
[2021-03-22 22:21:10] Computing distance between 10000 labelled and 50000 unlabelled vectors of length 512...
[2021-03-22 22:21:38] Searching for coresets with 20 beams...
[2021-03-22 22:26:53] Showing labelled: True (15000/50000 visible, 0 redundant)
[2021-03-22 22:26:53] Creating trainer with model on device: cuda
[2021-03-22 22:26:53] Training with expected score >= 0.9900
[2021-03-22 22:26:53] Training vgg-pretrained across 15000 data points in cifar100...
[2021-03-22 22:31:34] Training accuracy: 0.9975
[2021-03-22 22:31:34] Testing on 10000 data points...
[2021-03-22 22:31:36] Test score for 15000 training labels: 0.5798
[2021-03-22 22:31:36] Showing labelled: False (50000/50000 visible, 0 redundant)
[2021-03-22 22:31:36] Found 50000 unlabelled features.
[2021-03-22 22:31:50] Computing distance between 15000 labelled and 50000 unlabelled vectors of length 512...
[2021-03-22 22:32:36] Searching for coresets with 20 beams...
[2021-03-22 22:37:49] Showing labelled: True (20000/50000 visible, 0 redundant)
[2021-03-22 22:37:49] Creating trainer with model on device: cuda
[2021-03-22 22:37:49] Training with expected score >= 0.9900
[2021-03-22 22:37:49] Training vgg-pretrained across 20000 data points in cifar100...
[2021-03-22 22:44:04] Training accuracy: 0.9995
[2021-03-22 22:44:04] Testing on 10000 data points...
[2021-03-22 22:44:06] Test score for 20000 training labels: 0.6063
[2021-03-22 22:44:06] Experiment repeat 5/5
[2021-03-22 22:44:06] Seeded: 9
[2021-03-22 22:44:06] Using 10.00% labels of the dataset (5000/50000)
[2021-03-22 22:44:06] Randomly labelled 5000/50000
[2021-03-22 22:44:06] Showing labelled: True (5000/50000 visible, 0 redundant)
[2021-03-22 22:44:06] Seeded: 10
[2021-03-22 22:44:06] Running: experiment 0
[2021-03-22 22:44:06] Showing labelled: True (5000/50000 visible, 0 redundant)
[2021-03-22 22:44:06] Creating pretrained=True VGG16...
[2021-03-22 22:44:08] No parameter reset since we are using a pretrained model.
[2021-03-22 22:44:08] vgg-pretrained: initialized 15291300 parameters.
[2021-03-22 22:44:08] Creating trainer with model on device: cuda
[2021-03-22 22:44:08] Training with expected score >= 0.9900
[2021-03-22 22:44:08] Training vgg-pretrained across 5000 data points in cifar100...
[2021-03-22 22:49:14] Training accuracy: 0.9978
[2021-03-22 22:49:14] Testing on 10000 data points...
[2021-03-22 22:49:16] Test score for 5000 training labels: 0.4837
[2021-03-22 22:49:16] Showing labelled: False (50000/50000 visible, 0 redundant)
[2021-03-22 22:49:16] Found 50000 unlabelled features.
[2021-03-22 22:49:28] Computing distance between 5000 labelled and 50000 unlabelled vectors of length 512...
[2021-03-22 22:49:42] Searching for coresets with 20 beams...
[2021-03-22 22:54:56] Showing labelled: True (10000/50000 visible, 0 redundant)
[2021-03-22 22:54:56] Creating trainer with model on device: cuda
[2021-03-22 22:54:56] Training with expected score >= 0.9900
[2021-03-22 22:54:56] Training vgg-pretrained across 10000 data points in cifar100...
[2021-03-22 22:58:05] Training accuracy: 0.9959
[2021-03-22 22:58:05] Testing on 10000 data points...
[2021-03-22 22:58:07] Test score for 10000 training labels: 0.5416
[2021-03-22 22:58:07] Showing labelled: False (50000/50000 visible, 0 redundant)
[2021-03-22 22:58:07] Found 50000 unlabelled features.
[2021-03-22 22:58:20] Computing distance between 10000 labelled and 50000 unlabelled vectors of length 512...
[2021-03-22 22:58:51] Searching for coresets with 20 beams...
[2021-03-22 23:04:04] Showing labelled: True (15000/50000 visible, 0 redundant)
[2021-03-22 23:04:04] Creating trainer with model on device: cuda
[2021-03-22 23:04:04] Training with expected score >= 0.9900
[2021-03-22 23:04:04] Training vgg-pretrained across 15000 data points in cifar100...
[2021-03-22 23:08:46] Training accuracy: 0.9955
[2021-03-22 23:08:46] Testing on 10000 data points...
[2021-03-22 23:08:48] Test score for 15000 training labels: 0.5712
[2021-03-22 23:08:48] Showing labelled: False (50000/50000 visible, 0 redundant)
[2021-03-22 23:08:48] Found 50000 unlabelled features.
[2021-03-22 23:09:02] Computing distance between 15000 labelled and 50000 unlabelled vectors of length 512...
[2021-03-22 23:09:48] Searching for coresets with 20 beams...
[2021-03-22 23:15:02] Showing labelled: True (20000/50000 visible, 0 redundant)
[2021-03-22 23:15:02] Creating trainer with model on device: cuda
[2021-03-22 23:15:02] Training with expected score >= 0.9900
[2021-03-22 23:15:02] Training vgg-pretrained across 20000 data points in cifar100...
[2021-03-22 23:21:20] Training accuracy: 0.9981
[2021-03-22 23:21:20] Testing on 10000 data points...
[2021-03-22 23:21:22] Test score for 20000 training labels: 0.5978
[2021-03-22 23:21:22] Updated results: ../results/cifar100/main_lcbeam_redo/results.json
