Michael Nguyen
|
ba27a6d95d
|
added kusal preprocess code
|
2018-06-18 12:20:09 -05:00 |
Keith Ito
|
c7efb3c208
|
Merge pull request #5 from MycroftAI/amy-data
Trim leading and trailing silence
|
2018-04-02 13:04:32 -07:00 |
Keith Ito
|
567bf6c62e
|
Trim leading and trailing silence
|
2018-04-02 13:04:10 -07:00 |
Keith Ito
|
1579386764
|
Merge pull request #4 from MycroftAI/amy-data
Add preprocessor for Amy dataset
|
2018-04-02 12:47:50 -07:00 |
Keith Ito
|
9ac364a647
|
Add preprocessor for Amy data
|
2018-04-02 12:46:26 -07:00 |
Keith Ito
|
bdee55e2b9
|
Restore missing README
|
2018-03-19 23:33:50 -07:00 |
Keith Ito
|
93b73a4746
|
Merge pull request #3 from MycroftAI/tuning
Tuning for initial run
|
2018-03-18 21:26:20 -07:00 |
Keith Ito
|
736bf5c1bf
|
Use 50 Griffin-Lim iterations by default
|
2018-03-17 14:33:57 -07:00 |
Keith Ito
|
2a68ef4b27
|
Set use_cmudict to True by default
|
2018-03-17 14:31:42 -07:00 |
Keith Ito
|
122f9783e7
|
Drop preemphasis
|
2018-03-17 14:31:24 -07:00 |
Keith Ito
|
d0e09d064a
|
Bump default sample rate up to 22kHz
|
2018-03-17 14:28:58 -07:00 |
Keith Ito
|
475ed791bf
|
Limit frequencies used in mel spectrogram to [125, 7600].
|
2018-03-17 14:27:57 -07:00 |
Keith Ito
|
70ba4aaf69
|
Use vanilla exponential decay for learning rate
|
2018-03-17 14:24:38 -07:00 |
Keith Ito
|
1084310ed1
|
Use 512D embeddings
|
2018-03-17 14:22:39 -07:00 |
Keith Ito
|
14ee2615eb
|
Make embedding dimension configurable
|
2018-03-17 14:22:24 -07:00 |
Keith Ito
|
11a6a096d9
|
Merge pull request #2 from MycroftAI/attention
Use location-sensitive attention (hybrid)
|
2018-03-17 14:21:54 -07:00 |
Keith Ito
|
1ff4995eb4
|
Use location-sensitive attention (hybrid)
|
2018-03-17 14:08:09 -07:00 |
Keith Ito
|
14960104e5
|
Merge pull request #1 from MycroftAI/mycroft-changes
Apache license and remove non Apache-compatible code
|
2018-02-05 16:51:45 -08:00 |
Keith Ito
|
e35859d0b0
|
Remove dependency on inflect
|
2018-02-05 16:41:56 -08:00 |
Keith Ito
|
92b4379165
|
Remove dependency on unidecode
|
2018-02-05 16:41:56 -08:00 |
Keith Ito
|
fd83b876a4
|
Update README
|
2018-02-05 16:41:56 -08:00 |
Keith Ito
|
1f58edd56d
|
Apache 2 license
|
2018-02-05 14:28:23 -08:00 |
Keith Ito
|
8edcd55b3f
|
Merge pull request #106 from rafaelvalle/melspectrogram_offset
adding offset to melspectorgram
|
2018-01-18 18:42:17 -08:00 |
Rafael Valle
|
37f04cb0f1
|
adding offset to melspectorgram
|
2018-01-16 10:25:34 -08:00 |
Keith Ito
|
43762dc456
|
Update install instructions to mention TensorFlow 1.4
|
2017-11-08 11:47:20 -08:00 |
Keith Ito
|
08989cc355
|
Add a comment about the max_iters error
|
2017-10-16 09:35:43 -07:00 |
Keith Ito
|
6612fd3ef6
|
Merge pull request #61 from greenkeytech/tf1.4fix
fix for tensorflow 1.4.0
|
2017-10-14 09:28:03 -07:00 |
Matthew Goldey
|
7b09826434
|
fix for tensorflow 1.4.0
|
2017-10-13 14:56:00 -05:00 |
Keith Ito
|
268995c08c
|
Merge pull request #58 from r9y9/utf8
Use UTF-8 explicitly
|
2017-10-06 09:12:07 -07:00 |
Ryuichi Yamamoto
|
e198279ef2
|
Use UTF-8 explicitly
Fixes #57
|
2017-10-06 18:24:51 +09:00 |
Keith Ito
|
9fc433a870
|
Merge pull request #50 from keithito/trim-audio
Trim audio at the first silence
|
2017-09-26 09:53:14 -07:00 |
Keith Ito
|
d88640b7c0
|
Detect endpoint at first long silence
|
2017-09-26 09:51:22 -07:00 |
Keith Ito
|
e61fa839f9
|
Remove dead code
|
2017-09-26 08:46:21 -07:00 |
Keith Ito
|
faf7b8d083
|
Trim silence from output
Remove reduction of max_iters in eval and demo_server.
|
2017-09-25 10:03:03 -07:00 |
Keith Ito
|
522826dd77
|
Merge pull request #41 from keithito/tf-griffin-lim
Add TensorFlow implementation of Griffin-Lim
|
2017-09-12 20:58:42 -07:00 |
Keith Ito
|
2460969dc9
|
Clarify comment
|
2017-09-12 20:52:20 -07:00 |
Keith Ito
|
564c75d1ae
|
Remove TF from requirements and add to install instructions
|
2017-09-12 20:43:01 -07:00 |
Keith Ito
|
9190f660bc
|
Add TensorFlow Griffin-Lim implementation
Based on @kyubyong's code:
https://github.com/Kyubyong/tensorflow-exercises/blob/master/Audio_Processing.ipynb
Requires TF 1.3
|
2017-09-12 20:04:39 -07:00 |
Keith Ito
|
ab5dae1e18
|
Fix broken link
|
2017-09-07 10:52:47 -07:00 |
Keith Ito
|
b49b901b82
|
Add @MXGray's results to README
|
2017-09-06 21:38:41 -07:00 |
Keith Ito
|
9262f440b2
|
Clarification about passing hparams flag at eval time
|
2017-09-04 22:08:26 -07:00 |
Keith Ito
|
53a840c030
|
Merge pull request #40 from keithito/simplify-input
Make text processing simpler and more configurable
|
2017-09-04 21:55:46 -07:00 |
Keith Ito
|
c4e14ad3b9
|
Rename "pipeline" to "cleaners"
No need to introduce new terminology.
|
2017-09-04 21:54:23 -07:00 |
Keith Ito
|
9e1ea7a879
|
Clean up
|
2017-09-04 21:37:58 -07:00 |
Keith Ito
|
c724ac2f4c
|
Reference cleaner docs in TRAINING_DATA.md
|
2017-09-04 21:32:26 -07:00 |
Keith Ito
|
edd0925213
|
Use single quotes
|
2017-09-04 21:20:25 -07:00 |
Keith Ito
|
d49a938a71
|
Update documentation
|
2017-09-04 21:20:25 -07:00 |
Keith Ito
|
479976b6c5
|
Simplify text processing, make it easier to adapt to non-English data.
|
2017-09-04 20:04:59 -07:00 |
Keith Ito
|
0894cd9bf2
|
Make power a hyperparameter
|
2017-09-03 12:36:09 -07:00 |
Keith Ito
|
6a793ba8c4
|
Update model download instructions
Some versions of tar require the "j" flag to decompress
|
2017-08-22 17:42:26 -04:00 |