Tilman Kamp
6c6a4e08ca
Fix #1962
2019-03-18 18:53:04 +01:00
Tilman Kamp
238987a70f
Fix #1955
2019-03-14 12:25:08 +01:00
Reuben Morais
a7367d4b40
Make bin/run-ldc93s1.sh run faster
2019-03-13 10:18:16 -03:00
Quentin Brunet
179ba1b533
Fix import_cv2.py binary file + permissions
2019-03-04 13:30:30 +01:00
Nicolas Panel
881e55e842
Use latest TS dataset
...
see https://github.com/nicolaspanel/TrainingSpeech#last-releases--download for more info
2019-02-20 10:08:56 +01:00
Reuben Morais
c1212ffbb2
Fix shape of loaded preprocessed features
2019-02-12 20:31:24 -02:00
josh
7bee1cf445
train/test/dev checking now made explicit
2019-02-01 15:28:32 +01:00
josh
9810946035
usage
2019-02-01 13:36:19 +01:00
josh
e26aefc39b
better handling of empty TSV dir / comments
2019-02-01 13:32:27 +01:00
josh
e6e33eb3a7
import script for new CV
2019-02-01 00:38:02 +01:00
Nicolas Panel
08bea8480a
use TrainingSpeech release 2018-11-24_fr_FR
...
`2018-11-24_fr_FR` fixes many annotation issues (~500) such as empty audio etc.
2018-11-25 08:56:03 +01:00
kdavis-mozilla
81b16002b7
Fixed #1740 (Switchboard Importer Creates WAV's at 8KHz not 16KHz)
2018-11-24 08:22:38 +01:00
kdavis-mozilla
d5c9d3e88c
Fixed #1747 (Indicate ISIP transcript version)
2018-11-24 07:46:31 +01:00
Kelly Davis
50d62b8cf1
Merge pull request #1741 from robmsmt/master
...
changes to swb import
2018-11-23 16:08:14 +01:00
Rob
4efc5c6c4d
tidy pep8
2018-11-21 10:11:44 -05:00
Rob
e47e344a82
added downloaders and assert checks
2018-11-21 00:27:09 -05:00
Rob
8507f992c0
update with changes for swb
2018-11-19 22:19:39 -05:00
kdavis-mozilla
ba622892aa
Addressed review comment
2018-11-19 14:26:34 +01:00
kdavis-mozilla
6c903c8b19
Fixed #1735 (Fisher importer creates 32 not 16 bit wav's)
2018-11-18 17:41:00 +01:00
kdavis-mozilla
b2f967ac4d
Addressed review comments
2018-11-17 20:24:01 +01:00
kdavis-mozilla
f03669171f
Fixed #1726 (Imported 8khz training audio compromised by unfiltered upsampling)
2018-11-17 20:18:31 +01:00
Reuben Morais
d125acfb3d
Merge pull request #1696 from mozilla/remove-old-ctc
...
Remove old CTC decoder (Fixes #1675 )
2018-11-12 15:55:32 -02:00
Célian GARCIA
9fc6c3f59b
Fix log issue
...
Very minor change in order to log something like
``No directory "/home/<user>/src/mozilla/cv_corpus_v1" - extracting archive...`` instead of
``No directory "/home/<user>/src/mozilla/cv_corpus_v1.tar.gz"
2018-11-10 11:22:25 +01:00
Reuben Morais
a0333728eb
Build decoder package in automation
2018-11-08 18:35:42 -02:00
Alexandre Lissy
5d30afdbad
Training and directly exporting as TF Lite
2018-10-30 19:17:25 +01:00
Reuben Morais
f11ccbe39b
Merge pull request #1661 from mozilla/remove-init-frozen-model
...
Remove initialize_from_frozen_model flag and support code (Fixes #1659 )
2018-10-23 19:26:30 -03:00
Reuben Morais
46d1cece4f
Remove initialize_from_frozen_model flag and support code
2018-10-23 15:08:38 -03:00
Faissal Bensefia
529ebdfcb4
Fix import_cv.py
...
I have no idea how this ever worked
2018-10-19 23:23:07 +01:00
lissyx
5118f7ffed
Merge pull request #1647 from nicolaspanel/master
...
[TrainingSpeech importer] generate english-complatible dataset
2018-10-15 16:00:58 +02:00
nicolaspanel
61d9193ce5
[TrainingSpeech importer] add --english-compatible option
2018-10-15 10:18:14 +02:00
Nicolas Panel
609120f3da
[TrainingSpeech importer] stick to data/alphabet.txt
...
aims to generate english-complatible dataset since no FR alphabet.txt for now
see https://github.com/mozilla/DeepSpeech/pull/1599#issuecomment-426544379 for more info
2018-10-14 13:48:59 +02:00
kdavis-mozilla
c38dc099d3
Fixed #1638 (Update Hyperparameters for 0.3.0)
2018-10-12 17:26:32 +02:00
nicolaspanel
baf9624714
setup training_speech importer (see #1576 )
2018-10-11 14:00:40 +02:00
Alexandre Lissy
4daf1bf947
Add single shot inference test coverage
2018-10-02 14:20:45 +02:00
Alexandre Lissy
94b8c5c1a3
Remove deprecased tensorflow.contrib.learn.python.learn.datasets.base.maybe_download
...
Fixes #1531
Fixes #1477
2018-10-02 10:59:16 +02:00
Alexandre Lissy
3dad37eb05
Remove AOT
2018-09-20 17:02:36 +02:00
Tilman Kamp
db702817b0
Removing old training scripts and website support - #1534 #1538
2018-09-13 15:56:35 +02:00
Reuben Morais
51cac1031b
Address review comments
2018-08-03 14:46:05 -03:00
Reuben Morais
49490f5e68
Increase CI training epochs to guarantee overfitting
2018-08-02 13:22:24 -03:00
Reuben Morais
4e383ac59f
Add tool to convert graph protobuf to pbtxt
2018-08-02 13:22:24 -03:00
Reuben Morais
d2be00fad0
Add tool to find out which ops are needed by a graph
2018-08-02 13:22:24 -03:00
Reuben Morais
400c9c0e5d
Streaming inference
2018-08-02 13:22:24 -03:00
Alexandre Lissy
2d3818b492
Remove deprecated automation code
...
Fixes #1478
2018-07-30 12:36:31 +02:00
Alexandre Lissy
298d11cd76
Add test coverage for training from frozen model
2018-07-24 13:02:53 +02:00
Reuben Morais
6cafcda1dd
Use progressbar2 in import_cv.py
2018-07-11 13:10:49 -03:00
Alexandre Lissy
646c9178cf
Non positionnal arguments. Everywhere.
...
Fixes #1443
2018-07-04 17:12:52 +02:00
Scott Stevenson
ccd8f560ca
Remove keyword argument that has default value
...
When no `type` keyword argument is passed to the `add_argument` method
of an `argparse.ArgumentParser`, the type will be `str` by default and
therefore this does not need to be specified.
2018-05-29 18:28:59 +01:00
Alexandre Lissy
d7653d749b
Use LDC93S1 small language model
...
How to regenerate:
- Get a KenLM build
- $ (tr -d '[:digit:]|[:punct:]' < LDC93S1.txt | tr '[:upper:]' '[:lower:]'; head -n 500 ../lm/vocab.txt) > vocab.txt
- $ bin/lmplz --prune 0 0 1 --order 5 --text vocab.txt --arpa vocab.pruned.arpa
- $ bin/build_binary -s vocab.pruned.arpa vocab.pruned.lm
- $ generate_trie ../alphabet.txt vocab.pruned.lm vocab.txt vocab.trie
Fixes #1245
2018-02-15 14:56:02 +01:00
Alexandre Lissy
bcfd4fd409
Update for r1.5
2018-02-14 11:58:48 +01:00
Divyansh Jha
d378ac7ffa
reorder arguments in client.cc, client.py and client.js also update call sites in tc-test-utils.sh
2018-02-14 09:29:35 +05:30
