36 commits

Author	SHA1	Message	Date
Jean-Marc Valin	543ee94037	20-bit VQ	2019-02-15 15:13:14 -05:00
Jean-Marc Valin	b2940ed212	Use real features at the chunk edges rather than zeros	2019-01-24 14:16:30 -05:00
Jean-Marc Valin	b84a06dd08	Use a single u-law embedding	2019-01-21 16:52:57 -05:00
Jean-Marc Valin	dc082d7c1c	Making it easier to change the frame size	2019-01-21 15:16:11 -05:00
Jean-Marc Valin	38cd5cf08f	Remove useless (and possibly hurtful) residual connection ... I guess it's a bad idea to forward inputs directly	2019-01-17 23:17:42 -05:00
Jean-Marc Valin	8c271d60c4	Controlling per-gate sparsity	2018-12-10 16:15:50 -05:00
Jean-Marc Valin	d961d009a0	Managing to actually use sparse matrices ... Now 2x real-time!	2018-11-28 20:20:17 -05:00
Jean-Marc Valin	70fdf47471	Copyright headers	2018-11-25 13:15:19 -05:00
Jean-Marc Valin	477d08734d	Dump embedding	2018-11-23 23:33:35 -05:00
Jean-Marc Valin	d93239e955	Using non-cudnn version of the GRU for the weights ... Not sure how the layout of the CuDNN version is	2018-11-23 20:07:42 -05:00
Jean-Marc Valin	b0c61158f7	More meaningful names	2018-11-23 19:51:34 -05:00
Jean-Marc Valin	db7569c3da	moving hyper-parameters to new_lpcnet_model() arguments ... more cleaning up too	2018-10-24 14:57:34 -04:00
Jean-Marc Valin	4e331f377f	LPC generation from the cepstral coefficients	2018-10-21 02:45:21 -04:00
Jean-Marc Valin	fb1d4fdec2	...	2018-10-21 02:45:21 -04:00
Jean-Marc Valin	fa1d2824fa	Add diagonal to sparsity mask	2018-10-21 02:45:21 -04:00
Jean-Marc Valin	f13debcf65	Arbitrary 16x1 sparseness	2018-10-21 02:45:21 -04:00
Jean-Marc Valin	9756feefbd	...	2018-10-10 18:05:27 -04:00
Jean-Marc Valin	495f8ea5f3	second RNN	2018-10-10 18:05:27 -04:00
Jean-Marc Valin	03fa20d532	remove unused/dead code	2018-10-09 12:27:02 -04:00
Jean-Marc Valin	ea1391e174	deeper features	2018-10-03 22:30:44 -04:00
Jean-Marc Valin	639766b322	pitch embedding	2018-10-02 21:17:34 -04:00
Jean-Marc Valin	2d74d3189c	...	2018-10-02 18:26:42 -04:00
Jean-Marc Valin	c381db5688	Use excitation as input	2018-08-24 00:20:10 -04:00
Jean-Marc Valin	87cd75f6f4	Training seems to work	2018-08-16 22:43:13 -04:00
Jean-Marc Valin	785a2b2e84	Predicting pre-emphasized audio	2018-08-16 13:58:33 -04:00
Jean-Marc Valin	3d20cdaed4	Add prediction	2018-08-14 18:40:32 -04:00
Jean-Marc Valin	70789e6f43	audio-domain synthesis	2018-07-31 18:37:27 -04:00
Jean-Marc Valin	374ba430c4	stashing stuff here	2018-07-12 18:20:25 -04:00
Jean-Marc Valin	5d8a1313d6	decodes something...	2018-07-11 01:30:30 -04:00
Jean-Marc Valin	824dbecaec	decoder wip	2018-07-09 18:20:52 -04:00
Jean-Marc Valin	06511ba5a4	Add convolution	2018-06-26 22:52:24 -04:00
Jean-Marc Valin	f884d7bf18	add convolution	2018-06-26 17:14:29 -04:00
Jean-Marc Valin	fd9002e98e	Adding pitch	2018-06-26 16:40:55 -04:00
Jean-Marc Valin	617e462be3	using features (except pitch gain which has NaNs for now)	2018-06-26 01:31:44 -04:00
Jean-Marc Valin	b65031ef64	excitation model	2018-06-25 16:26:47 -04:00
Jean-Marc Valin	c41afe41f0	initial commit	2018-06-21 20:45:54 -04:00