3882334 | Luke Yeager | 18 February 2016, 00:12:23 UTC | Mark v3.2.0 | 18 February 2016, 00:12:23 UTC |
340e42d | Luke Yeager | 17 February 2016, 19:04:56 UTC | Merge pull request #582 from anuragphadke/master Classification example np.array performance bottleneck | 17 February 2016, 19:04:56 UTC |
843848a | Anurag Phadke | 17 February 2016, 18:43:19 UTC | https://github.com/NVIDIA/DIGITS/issues/581 | 17 February 2016, 18:43:19 UTC |
1cb4de2 | Luke Yeager | 17 February 2016, 18:04:33 UTC | Merge pull request #580 from anuragphadke/master pass batch_size to classify() via command line | 17 February 2016, 18:04:33 UTC |
5f7217c | Anurag Phadke | 17 February 2016, 15:07:21 UTC | pass batch_size to classify() via command line | 17 February 2016, 15:07:21 UTC |
7f2f820 | Luke Yeager | 17 February 2016, 00:11:45 UTC | Merge pull request #579 from lukeyeager/classification-example-timing Fix timing in example/classification scripts | 17 February 2016, 00:11:45 UTC |
7f4f1fb | Luke Yeager | 16 February 2016, 21:18:31 UTC | Fix timing in example/classification scripts | 16 February 2016, 21:18:31 UTC |
9419536 | Greg Heinrich | 14 February 2016, 18:37:27 UTC | Merge pull request #575 from gheinrich/dev/bug-574 Fix name of ip2_right bias | 14 February 2016, 18:37:27 UTC |
e538b51 | Luke Yeager | 12 February 2016, 17:37:49 UTC | Merge pull request #576 from gheinrich/dev/caffe-net-sanity-check Drop error if a Caffe layer references an invalid bottom | 12 February 2016, 17:37:49 UTC |
d3cbdca | Greg Heinrich | 11 February 2016, 15:17:09 UTC | Drop error if a Caffe layer references an invalid bottom This should help address the case when a layer is added to manipulate e.g. the label blob. In that case the layer must include a directive like: include { phase: TRAIN } Otherwise Caffe will drop an error during inference. A sanity check is added to DIGITS to catch these errors earlier in the process (i.e. before training). | 11 February 2016, 21:17:41 UTC |
ada2033 | Greg Heinrich | 11 February 2016, 13:14:42 UTC | Fix name of ip2_right bias closes #574 | 11 February 2016, 13:14:42 UTC |
17668e3 | Luke Yeager | 08 February 2016, 18:54:57 UTC | Merge pull request #571 from lukeyeager/customize-float-snapshot Cast to float (not int) when customizing snapshots | 08 February 2016, 18:54:57 UTC |
7310dfc | Luke Yeager | 08 February 2016, 18:50:49 UTC | Merge pull request #541 from lukeyeager/single-config-file Only look in one location for config files | 08 February 2016, 18:50:49 UTC |
dfeb185 | Luke Yeager | 08 February 2016, 18:19:56 UTC | Cast to float (not int) when customizing snapshots | 08 February 2016, 18:19:56 UTC |
9e08684 | Luke Yeager | 04 February 2016, 22:54:50 UTC | Merge pull request #569 from dongjoon-hyun/fix_illformed_html Fix an ill-formed html file consistently with other html files. | 04 February 2016, 22:54:50 UTC |
dea205e | Dongjoon Hyun | 04 February 2016, 22:43:41 UTC | Fix an ill-formed html file consistently with other html files. `<ul>` tag should be outside of `<p>` tag. This PR fixes `classification/custom_network_explanation.html`. The result is consistent with `generic/customer_network_explanation.html` and others. | 04 February 2016, 22:45:02 UTC |
04655eb | Luke Yeager | 04 February 2016, 17:57:46 UTC | Merge pull request #568 from gheinrich/dev/fix-job-detail-link Fix links to jobs on job details page | 04 February 2016, 17:57:46 UTC |
2211e54 | Greg Heinrich | 04 February 2016, 14:04:35 UTC | Fix link to jobs on job details page closes #567 | 04 February 2016, 14:04:35 UTC |
ebfb188 | Luke Yeager | 03 February 2016, 21:21:27 UTC | Merge pull request #564 from lukeyeager/add-new-solvers Add RMSprop, AdaDelta and Adam solvers | 03 February 2016, 21:21:27 UTC |
15cade9 | Luke Yeager | 03 February 2016, 17:54:59 UTC | Add new Torch solver types | 03 February 2016, 17:54:59 UTC |
362ebe9 | Luke Yeager | 02 February 2016, 23:26:26 UTC | Add RMSprop, AdaDelta and Adam solvers Disable solvers which aren't supported by the selected framework | 03 February 2016, 00:14:13 UTC |
ea2a941 | Greg Heinrich | 02 February 2016, 21:13:12 UTC | Merge pull request #561 from dongjoon-hyun/fix_unmatched_assignment Fix unmatched local variable initialization. | 02 February 2016, 21:13:12 UTC |
3588f5a | Luke Yeager | 02 February 2016, 19:26:48 UTC | Merge pull request #563 from lukeyeager/install-cuda-doc Use echo/tee to create file at /etc/ld.so.conf.d/ | 02 February 2016, 19:26:48 UTC |
32be9b1 | Luke Yeager | 02 February 2016, 18:44:33 UTC | Use echo/tee to create file at /etc/ld.so.conf.d/ | 02 February 2016, 18:45:21 UTC |
db9729f | Luke Yeager | 02 February 2016, 18:36:43 UTC | Merge pull request #558 from lukeyeager/install-cuda-doc Add documentation for installing CUDA | 02 February 2016, 18:36:43 UTC |
4ee80d1 | Luke Yeager | 02 February 2016, 18:31:09 UTC | Use "Deb" when referring to packages in docs | 02 February 2016, 18:31:09 UTC |
7367bb6 | Luke Yeager | 29 January 2016, 18:59:22 UTC | Add documentation for installing CUDA | 02 February 2016, 18:29:24 UTC |
cb0cba0 | Luke Yeager | 02 February 2016, 17:42:16 UTC | Merge pull request #557 from lukeyeager/classification-example-gotchas Address gotchas with the classification example | 02 February 2016, 17:42:16 UTC |
09b54b0 | Luke Yeager | 02 February 2016, 17:03:43 UTC | Merge pull request #562 from lukeyeager/luarocks-lightningmdb Revert "Use last version of lightningmdb wrapper for Torch" | 02 February 2016, 17:03:43 UTC |
79c3651 | Dongjoon Hyun | 02 February 2016, 16:53:46 UTC | According to the comments, declare local variable `v`. | 02 February 2016, 16:53:46 UTC |
fbd2903 | Luke Yeager | 01 February 2016, 18:07:49 UTC | Merge pull request #559 from lukeyeager/classification-example-improvements Improvements to examples/classification/ | 01 February 2016, 18:07:49 UTC |
66b9b40 | Dongjoon Hyun | 01 February 2016, 17:51:25 UTC | Fix unmatched local variable initialization. | 01 February 2016, 17:51:25 UTC |
4ba871f | Luke Yeager | 01 February 2016, 17:45:55 UTC | Merge pull request #560 from kumadasu/fix-broken-link-of-document Fix broken link to UbuntuInstall.md | 01 February 2016, 17:45:55 UTC |
a8222cc | Luke Yeager | 01 February 2016, 17:42:49 UTC | Revert "Use last version of lightningmdb wrapper for Torch" This reverts commit 3afb6ab5017a234a02e47629a27a8bacddded3f7. | 01 February 2016, 17:42:49 UTC |
e5b05cc | kumadasu | 31 January 2016, 08:49:30 UTC | Fix broken link to UbuntuInstall.md | 31 January 2016, 08:49:30 UTC |
abf719b | Luke Yeager | 29 January 2016, 22:10:34 UTC | Improvements to examples/classification/ Allow multiple image files Add batch_size | 29 January 2016, 22:10:34 UTC |
8d9a3b5 | Luke Yeager | 29 January 2016, 21:29:42 UTC | Address gotchas with the classification example | 29 January 2016, 21:31:44 UTC |
1b885fb | Greg Heinrich | 28 January 2016, 16:27:03 UTC | Merge pull request #552 from gheinrich/dev/regression-model-input-check Make Torch regression example more robust | 28 January 2016, 16:27:03 UTC |
4ac4680 | Luke Yeager | 27 January 2016, 19:32:15 UTC | Merge pull request #551 from lukeyeager/fix-loading-bad-jobs Fix error-handling on loading jobs with no job_id | 27 January 2016, 19:32:15 UTC |
533f59b | Luke Yeager | 27 January 2016, 18:55:57 UTC | Fix error-handling on loading jobs with no job_id | 27 January 2016, 18:55:57 UTC |
d681946 | Luke Yeager | 27 January 2016, 17:15:55 UTC | Merge pull request #550 from gheinrich/dev/torch-multistep-lr-policy Fix Torch handling of multistep LR policy | 27 January 2016, 17:15:55 UTC |
3c41e05 | Greg Heinrich | 27 January 2016, 08:47:31 UTC | Fix Torch handling of multistep LR policy fix #547 | 27 January 2016, 08:47:31 UTC |
5c1372c | Luke Yeager | 27 January 2016, 00:39:18 UTC | Merge pull request #549 from lukeyeager/fix-multistep Fix multistep learning rate parsing | 27 January 2016, 00:39:18 UTC |
709d5b1 | Luke Yeager | 26 January 2016, 23:49:39 UTC | Merge pull request #548 from lukeyeager/python-layer-multigpu-note Add warning about Python layers and multi-GPU (master) | 26 January 2016, 23:49:39 UTC |
8962ddd | Luke Yeager | 26 January 2016, 23:46:18 UTC | Fix multistep learning rate parsing | 26 January 2016, 23:47:04 UTC |
2cd401f | Luke Yeager | 26 January 2016, 23:00:49 UTC | Add warning about Python layers and multi-GPU | 26 January 2016, 23:00:49 UTC |
a95b89e | Luke Yeager | 26 January 2016, 21:41:11 UTC | Merge pull request #543 from lukeyeager/travis-scripts-without-travis Use Travis scripts outside of Travis more easily | 26 January 2016, 21:41:11 UTC |
5026722 | Luke Yeager | 26 January 2016, 01:51:24 UTC | Make install-caffe script work outside Travis env | 26 January 2016, 20:55:52 UTC |
381bba9 | Luke Yeager | 26 January 2016, 01:51:11 UTC | Remove unnecessary pushd statement | 26 January 2016, 20:55:52 UTC |
79eb08e | Luke Yeager | 25 January 2016, 19:42:34 UTC | Only look in one location for config files Look at digits/digits.cfg Don't look at $HOME/.digits.cfg or /etc/digits.cfg | 26 January 2016, 20:55:32 UTC |
baa929a | Luke Yeager | 26 January 2016, 20:53:59 UTC | Merge pull request #545 from gheinrich/dev/torch-lmdb-package Use last version of lightningmdb wrapper for Torch | 26 January 2016, 20:53:59 UTC |
3afb6ab | Greg Heinrich | 26 January 2016, 18:30:01 UTC | Use last version of lightningmdb wrapper for Torch closes #544 | 26 January 2016, 18:30:52 UTC |
07a4697 | Luke Yeager | 26 January 2016, 17:37:48 UTC | Merge pull request #542 from lukeyeager/torch-nccl-instructions Install torch.nccl luarocks pkg w/ single command | 26 January 2016, 17:37:48 UTC |
2af845d | Luke Yeager | 26 January 2016, 17:36:50 UTC | Install torch.nccl luarocks pkg w/ single command | 26 January 2016, 17:36:50 UTC |
e8bfa4d | Greg Heinrich | 26 January 2016, 15:48:02 UTC | Make Torch regression example more robust Check if `inputShape` field of external parameters is not nil before calculating dimension of input tensor. Set dimension to `1` by default for network visualization. | 26 January 2016, 16:14:44 UTC |
dc9f827 | Luke Yeager | 22 January 2016, 18:54:43 UTC | Merge pull request #538 from gheinrich/dev/bug-533 Re-use all weights when re-training on same dataset | 22 January 2016, 18:54:43 UTC |
a29ad9a | Luke Yeager | 22 January 2016, 18:33:18 UTC | Merge pull request #539 from gheinrich/dev/alexnet-on-more-than-2-gpus Allow standard Alexnet model for Torch to train on more than 2 GPUs | 22 January 2016, 18:33:18 UTC |
95b9254 | Greg Heinrich | 22 January 2016, 14:59:34 UTC | Re-use all weights when re-training on same dataset Current behaviour: When a classification model is being re-trained the last `InnerProduct` layer is renamed to allow training on another dataset (with a possibly different number of classes). This implies that the weights of that last layer are discarded. Problem: If the model is being re-trained on the same dataset then initially the model will perform badly as it needs to re-learn the weights of its last `InnerProduct` layer. Solution: Refrain from renaming the last `InnerProduct` layer if the model is being re-trained on the same dataset Note: This applies to Caffe only. closes #533 | 22 January 2016, 18:18:09 UTC |
22461de | Luke Yeager | 22 January 2016, 17:45:12 UTC | Mark v3.2-dev Signifies that we're starting development on v3.2 | 22 January 2016, 17:45:12 UTC |
3845aa5 | Greg Heinrich | 22 January 2016, 16:07:52 UTC | Allow standard Alexnet model for Torch to train on more than 2 GPUs | 22 January 2016, 16:07:52 UTC |
9718fbe | Luke Yeager | 22 January 2016, 00:07:14 UTC | Mark v3.1.0 | 22 January 2016, 00:07:42 UTC |
d4ce486 | Luke Yeager | 22 January 2016, 00:05:52 UTC | Merge pull request #537 from lukeyeager/docs-3.1 Update docs for v3.1 | 22 January 2016, 00:05:52 UTC |
92a7c05 | Luke Yeager | 21 January 2016, 23:02:50 UTC | Update docs for v3.1 | 21 January 2016, 23:05:31 UTC |
f04e7ba | Luke Yeager | 21 January 2016, 00:09:49 UTC | Add links to examples/ in GettingStarted.md | 21 January 2016, 23:05:30 UTC |
8711181 | Luke Yeager | 21 January 2016, 16:54:22 UTC | Merge pull request #534 from gheinrich/dev/bug-532 Allow non-integer step size in visualize_lr() | 21 January 2016, 16:54:22 UTC |
da06d42 | Greg Heinrich | 21 January 2016, 10:09:56 UTC | Allow non-integer step size in visualize_lr() closes #532 | 21 January 2016, 10:09:56 UTC |
301b287 | Luke Yeager | 20 January 2016, 22:00:42 UTC | Merge pull request #531 from lukeyeager/fix-nginx-redirection Fix nginx redirection to ports other than 80 (master) | 20 January 2016, 22:00:42 UTC |
7c2ebd9 | Luke Yeager | 20 January 2016, 20:59:52 UTC | Fix nginx redirection to ports other than 80 | 20 January 2016, 20:59:52 UTC |
b9bd74d | Luke Yeager | 20 January 2016, 19:55:33 UTC | Merge pull request #529 from dongjoon-hyun/add_missing_classmethod_decorator Add a missing `classmethod` decorator to `prompt_message` function. | 20 January 2016, 19:55:33 UTC |
a3850b8 | Dongjoon Hyun | 20 January 2016, 18:41:37 UTC | Add a missing `classmethod` decorator to `prompt_message` function. | 20 January 2016, 18:41:37 UTC |
f62d624 | Luke Yeager | 19 January 2016, 20:41:20 UTC | Merge pull request #500 from gheinrich/dev/fine-tuning-tutorial Fine-tuning tutorial | 19 January 2016, 20:41:20 UTC |
65f41ef | Luke Yeager | 19 January 2016, 20:41:06 UTC | Merge pull request #499 from gheinrich/dev/torch-fine-tuning Torch model fine-tuning | 19 January 2016, 20:41:06 UTC |
1caa9dd | Luke Yeager | 19 January 2016, 20:40:57 UTC | Merge pull request #526 from gheinrich/dev/image-folder-in-inference Allow optional specification of image folder during multiple inference | 19 January 2016, 20:40:57 UTC |
a788863 | Luke Yeager | 19 January 2016, 20:40:46 UTC | Merge pull request #527 from dongjoon-hyun/add_missing_html_tag Add a missing `dd` tag and fix an unmatched `dt` tag | 19 January 2016, 20:40:46 UTC |
abf81f1 | Greg Heinrich | 19 January 2016, 15:30:59 UTC | Allow optional specification of image folder during multiple inference closes #504 | 19 January 2016, 20:04:58 UTC |
dba804e | Dongjoon Hyun | 19 January 2016, 18:26:52 UTC | Add a missing `dd` tag and fix an unmatched `dt` tag in templates/datasets/images. | 19 January 2016, 18:31:10 UTC |
3845226 | Luke Yeager | 19 January 2016, 17:28:15 UTC | Merge pull request #523 from dongjoon-hyun/use_cls_for_classmethod Use `cls` for classmethod instead of `self`. | 19 January 2016, 17:28:15 UTC |
1d768cc | Luke Yeager | 19 January 2016, 17:13:00 UTC | Merge pull request #525 from gheinrich/dev/fix-import Fix another import bug from #501 | 19 January 2016, 17:13:00 UTC |
56bfae8 | Greg Heinrich | 19 January 2016, 12:43:54 UTC | Fix another import bug from #501 | 19 January 2016, 15:32:13 UTC |
10ae024 | Dongjoon Hyun | 19 January 2016, 05:21:51 UTC | Use `cls` for classmethod instead of `self`. | 19 January 2016, 05:21:51 UTC |
845973f | Luke Yeager | 18 January 2016, 20:42:42 UTC | Merge pull request #522 from gheinrich/dev/weight-init-example Weight initialization example | 18 January 2016, 20:42:42 UTC |
729f0b7 | Greg Heinrich | 10 January 2016, 22:03:46 UTC | Torch model fine-tuning | 18 January 2016, 20:35:36 UTC |
ebd0da7 | Luke Yeager | 18 January 2016, 18:52:33 UTC | Merge pull request #498 from IgorX2/master Add option to use multiple pretrained networks | 18 January 2016, 18:52:33 UTC |
66d07c6 | Greg Heinrich | 18 January 2016, 16:03:27 UTC | Weight initialization example | 18 January 2016, 17:54:35 UTC |
5fddcec | Luke Yeager | 18 January 2016, 17:34:37 UTC | Merge pull request #519 from gheinrich/dev/siamese-tuto-create-db-fix Siamese DB creation fixes | 18 January 2016, 17:34:37 UTC |
be42593 | Greg Heinrich | 10 January 2016, 22:07:12 UTC | Fine-tuning tutorial | 18 January 2016, 17:25:14 UTC |
1cc36ec | Luke Yeager | 18 January 2016, 17:17:24 UTC | Merge pull request #520 from dongjoon-hyun/fix_typos_in_tools Fix typos in tools | 18 January 2016, 17:17:24 UTC |
b8b03e2 | Luke Yeager | 18 January 2016, 17:16:39 UTC | Merge pull request #486 from gheinrich/dev/siamese-tutorial-torch Siamese network tutorial for Torch | 18 January 2016, 17:16:39 UTC |
f08cfb0 | Greg Heinrich | 23 December 2015, 10:54:28 UTC | Siamese network tutorial for Torch | 18 January 2016, 10:31:39 UTC |
5d02951 | Dongjoon Hyun | 18 January 2016, 07:21:57 UTC | Fix typos in tools | 18 January 2016, 07:25:09 UTC |
25eec09 | Greg Heinrich | 17 January 2016, 18:22:12 UTC | Siamese tutorial create_db mean fix Remove extra channel transpose - already done in _save_image() Fixes one issue reported in #515 | 17 January 2016, 22:04:56 UTC |
5969172 | Greg Heinrich | 17 January 2016, 22:02:24 UTC | Siamese tutorial DB batch writing Enable batch writing in order to speed up DB creation. fix slowliness issue reported in #515 and #486 | 17 January 2016, 22:02:24 UTC |
a63e129 | Igor Ševo | 16 January 2016, 16:04:25 UTC | Add delimiter space stripping to pretrained models | 16 January 2016, 16:04:25 UTC |
60c0cad | Luke Yeager | 15 January 2016, 23:17:37 UTC | Merge pull request #485 from gheinrich/dev/torch-table-outputs Support table of outputs in Torch models | 15 January 2016, 23:17:37 UTC |
4116b5a | Luke Yeager | 15 January 2016, 22:42:45 UTC | Merge pull request #517 from lukeyeager/analyze-without-shape Allow analyze_db tool to read Datums without shape | 15 January 2016, 22:42:45 UTC |
800cb0f | Luke Yeager | 15 January 2016, 21:51:19 UTC | Merge pull request #516 from lukeyeager/stringio-imports Standardize cStringIO/StringIO imports | 15 January 2016, 21:51:19 UTC |
2d633ee | Luke Yeager | 15 January 2016, 17:59:13 UTC | Allow analyze_db tool to read Datums without shape | 15 January 2016, 21:03:01 UTC |
fa56ee5 | Luke Yeager | 15 January 2016, 20:54:50 UTC | Change the way we mock TestLoadImage.test_good_url Less mocking is good. Also I couldn't figure out how to mock StringIO anymore. | 15 January 2016, 20:54:50 UTC |
787787d | Luke Yeager | 15 January 2016, 17:38:47 UTC | Standardize cStringIO/StringIO imports | 15 January 2016, 17:38:47 UTC |
7b94e80 | Greg Heinrich | 14 January 2016, 20:47:07 UTC | Merge pull request #507 from lukeyeager/flask-blueprints Use flask blueprints to cleanup routing code | 14 January 2016, 20:47:07 UTC |