Commit Graph

53 Commits

Author SHA1 Message Date
Kelly Davis
aad1e99784 Merge pull request #431 from mozilla/issue406
Fixed #406
2017-03-10 11:33:23 +01:00
Kelly Davis
de45b40d8e Fixed #406 2017-03-10 07:27:20 +01:00
Kelly Davis
7fa0238dac Fixed #428 2017-03-08 17:24:14 +01:00
Kelly Davis
3b840625fd Fixed #418 2017-03-06 08:06:44 +01:00
Kelly Davis
b1ce790e4e Fix issue #395 2017-02-22 16:41:45 +01:00
Alexandre Lissy
2f63b2e89a Move some data to be stored on /data
Settting temp directory to be in /data, and DATA_DIR to /data,
defaulting to XDG_DATA_HOME.

Fixes #374
2017-02-13 14:25:58 +01:00
Kelly Davis
94aa5ba667 Fix #368 2017-02-10 12:46:39 +01:00
Kelly Davis
a8c473d00b Fix #357 2017-02-05 08:18:13 +01:00
Reuben Morais
2c832169de Merge pull request #352 from mozilla/issue313_update
Notebook decomposition PR, updated
2017-02-03 09:17:59 -02:00
Reuben Morais
7f5420c337 Convert notebook to script 2017-01-28 18:19:45 -02:00
Kelly Davis
c5211d84c3 Fix #349 2017-01-28 08:15:11 +01:00
Kelly Davis
67c48d252a Fixed #346 2017-01-23 06:24:23 +01:00
Kelly Davis
dffaa95a23 Fix #336 2017-01-07 09:31:50 +01:00
Andre Natal
d8ff6fcc64 Adding script to run swb 2017-01-03 12:17:02 -08:00
Reuben Morais
bc8b046605 Merge pull request #84 from mozilla/issue3_fisher
Implement Fisher corpus importer
2016-12-30 09:29:36 -02:00
Reuben Morais
ec590bd40b Remove ds_dataset_path export from run-fisher.sh 2016-12-30 09:28:37 -02:00
Kelly Davis
f2a1df681a Fix #325 (Dropout too low) 2016-12-29 15:47:25 +01:00
Reuben Morais
53f76b30f8 Add bin/run-fisher.sh script for training with Fisher 2016-12-28 12:44:24 -02:00
Kelly Davis
3533852884 Decreased batch sizes 2016-12-24 08:07:40 +01:00
Kelly Davis
7492eb2f8c Decreased batch sizes 2016-12-24 08:07:08 +01:00
Kelly Davis
a6ef802780 Fixed #311 Increased model capacity 2016-12-23 15:51:19 +01:00
Kelly Davis
9658061840 Fixed #309 2016-12-22 06:20:33 +01:00
Kelly Davis
fa283cff0d Fixed #275 2016-12-04 18:26:20 +01:00
Kelly Davis
a65beddb28 Updated learning_rate to fix inf's 2016-12-01 15:05:24 +01:00
Kelly Davis
7d3aff6131 Updated in line with comments 2016-11-30 22:30:17 +01:00
Chris Lord
85cd407a18 Rename 'training_iters' -> 'epochs' 2016-11-30 18:05:30 +00:00
Chris Lord
ffc76343dc Add a README for the binaries directory 2016-11-30 18:02:36 +00:00
Chris Lord
fce2cacf19 Tweak scripts to more appropriate default values 2016-11-30 17:58:36 +00:00
Kelly Davis
ad735d63ae Fix issue #259 2016-11-30 06:45:09 +01:00
Kelly Davis
19ade3ef44 Merge pull request #245 from mozilla/issue244
Fixed #244
2016-11-28 12:09:28 +01:00
Kelly Davis
077c4cda33 Fixed #244 2016-11-28 10:56:00 +01:00
Kelly Davis
06bb017c15 Fixed #239 2016-11-28 09:44:41 +01:00
Kelly Davis
002f0dd505 Made the system not validate during wer-tracking 2016-11-26 07:56:42 +01:00
Kelly Davis
79b41afa49 Fixed #229 2016-11-25 20:31:55 +01:00
Kelly Davis
2af3bc68f2 Fix #221 2016-11-18 17:33:13 +01:00
Kelly Davis
4454336d50 Merge pull request #213 from mozilla/issue201
Issue201
2016-11-17 19:24:47 +01:00
Kelly Davis
1da578f04c Updated params inline with IRC discussion 2016-11-17 18:09:54 +01:00
Alexandre Lissy
58f0609afa Tracking GPU Usage
Fixes #178
2016-11-17 11:49:20 +01:00
Kelly Davis
27f2c5f90f 64 leads to OOM 2016-11-17 06:31:46 +01:00
Kelly Davis
1766ba808b Fixed #201 2016-11-16 22:37:11 +01:00
Alexandre Lissy
2c18cb25fb Forcing python to not buffer output
Fixes #198
2016-11-16 13:58:35 +01:00
Chris Lord
348a2d0f72 Merge pull request #179 from Cwiiis/export-on-automation
Export the model on automation, fixes #177
2016-11-11 11:50:40 +01:00
Chris Lord
fe7d618b23 Export the model on automation 2016-11-11 11:35:39 +01:00
Kelly Davis
418e9e03a7 Made dirs server centric 2016-11-11 11:29:11 +01:00
Kelly Davis
39e7d5e7c6 Fixed typo 2016-11-11 07:10:21 +01:00
Kelly Davis
73330104c8 Fixed #174 2016-11-11 07:09:00 +01:00
Alexandre Lissy
b3f4f40869 Do not remove existing checkpoint dirs, update training paramteters
This breaks restore feature. We should let code overwrite existing
checkpoints. This fixes #166.

Change training parameters, to address issue #169.
2016-11-10 15:54:09 +01:00
Alexandre Lissy
06f28af274 Limit training epochs to 6
Fixes #160
2016-11-10 11:13:00 +01:00
Alexandre Lissy
1907f9eead Change automation parameters
Set importer to 'ted' instead of 'ted_lium', fixes #149
Set batch_size to 32, fixes #150
2016-11-09 13:04:46 +01:00
Alexandre Lissy
1150c1c5ec Fix TED LIUM automation path
Fixes #96
2016-10-26 15:53:40 +02:00