81 Commits

Author	SHA1	Message	Date
Eren Golge	e41a7c1a02	setup gradual training schedule wrt num gpus	2019-11-20 12:08:23 +01:00
Eren Golge	015f7780f4	Decoder shape comments for Tacotron2, decoupled grad clip for stopnet and the rest of the network. Some variable renaming and bug fix for alignment score logging	2019-11-12 11:20:53 +01:00
Eren Golge	adf9ebd629	Graves attention and setting attn type by config.json	2019-11-12 11:18:57 +01:00
Eren Golge	e83a4b07d2	commention model outputs for tacotron, align outputs shapes of tacotron and tracotron2, merge bidirectional decoder	2019-10-28 14:51:19 +01:00
Eren Golge	acbafb456b	Weighting positive values for stopnet loss, change adam_weight_decay name	2019-09-28 15:44:17 +02:00
Eren Golge	99d7f2a666	update set_weight_decay	2019-09-28 15:31:18 +02:00
Eren Golge	8565c508e4	remove debug line	2019-09-28 01:11:04 +02:00
Eren Golge	b76aaf8ad4	skip weight decay for BN and biases, some formatting	2019-09-28 01:09:28 +02:00
Eren Golge	5b6b1f354d	add use_gst to enable global style token	2019-09-24 16:24:58 +02:00
Eren Golge	d45d963dc1	linter fix	2019-09-11 10:39:59 +02:00
Eren Golge	609d8efa69	compute alignment diagonality score and encapsulate stats averaging with a class in traning	2019-09-11 10:32:07 +02:00
Reuben Morais	28644a717e	Fix tests	2019-08-29 12:18:33 +02:00
Eren Golge	b22c7d4a29	Merge branch 'dev-gradual-queue' into dev	2019-08-16 13:20:17 +02:00
Eren Golge	5acd9e82bd	save model r value for checkpoints	2019-08-16 13:11:51 +02:00
Eren Golge	91795cc0f1	config update	2019-07-22 15:44:09 +02:00
Eren Golge	ee706b50f6	enalbe graudal training by config.json	2019-07-22 02:11:20 +02:00
Reuben Morais	11e7895329	Fix Pylint issues	2019-07-19 09:08:51 +02:00
Eren Golge	fd081c49b7	split dataset outside preprocessor	2019-07-16 21:15:04 +02:00
Eren Golge	aec7f02817	libri tts config, and bug fix	2019-07-16 15:17:38 +02:00
Thomas Werkmeister	2f2482f9b4	reading all speakers upfront	2019-07-10 18:38:55 +02:00
Thomas Werkmeister	d172a3d3d5	multispeaker	2019-06-26 12:59:14 +02:00
Eren Golge	7410daceb2	Adapt TTS for TacotronGST and some changes for Audio.py , better config.json naming	2019-06-05 18:33:57 +02:00
Eren Golge	4678c66599	forward_attn_mask and config update	2019-06-04 00:39:29 +02:00
Eren Golge	ba492f43be	Set tacotron model parameters to adap to common_layers.py - Prenet and Attention	2019-05-27 14:40:28 +02:00
Eren Golge	e62659da94	update separate stopnet flow to make it faster.	2019-05-17 16:15:43 +02:00
Eren Golge	832dc3eafa	bug fix	2019-05-15 12:37:31 +02:00
Eren Golge	6331bccefc	make dropout oprional #2	2019-05-12 17:35:31 +02:00
Eren Golge	e2439fde9a	make location attention optional and keep all attention weights in attention class	2019-04-29 11:37:01 +02:00
Eren Golge	3c2d500f53	Changesat windowing and some comments	2019-04-12 16:13:40 +02:00
Eren Golge	e2cf35bb10	Make loss masking optional	2019-04-10 16:41:08 +02:00
Eren Golge	8a47b46195	print warning if a layer in ehckpoint is not defined in model definition	2019-04-08 19:32:07 +02:00
Eren Golge	961af0f5cd	setup_model externally based on model selection. Make forward attention and prenet type configurable in config.json	2019-04-05 17:49:18 +02:00
Eren Golge	6edd8bc6dd	add git branch and restore_path to copied config file for each run	2019-03-29 17:01:57 +01:00
Eren Golge	d8908692c5	refactor partial reinit script as a function. Allow user to select layers to reinit in finutunning	2019-03-23 17:19:40 +01:00
Eren Golge	08162157ee	generic train.py for multiple architectures set on config.json	2019-03-06 13:11:22 +01:00
Eren Golge	bf5f18d11e	Formatting changes and distributed training	2019-02-27 09:50:52 +01:00
Eren Golge	caae1af4f6	visual updates for phoenemes	2019-02-25 17:20:36 +01:00
Eren Golge	806643300c	Place model name to the beginning of the generated output folder name	2018-12-28 14:22:41 +01:00
Eren Golge	bb2a88a984	Rename LR scheduler	2018-11-26 14:09:42 +01:00
Eren Golge	22dcc4f7d0	small print formatting	2018-11-22 17:03:53 +01:00
Eren Golge	6550db5251	Formatting, fixing import statements, logging learning rate, remove optimizer restore cuda call	2018-11-05 14:05:04 +01:00
Eren Golge	0b6a9995fc	change import statements	2018-11-03 19:15:06 +01:00
Eren Golge	c8a552e627	Batch update after data-loss	2018-11-02 16:13:51 +01:00
Eren	bb526c296f	Change scheduler AnnealLR and catch audio synthesis error in eval time	2018-08-13 13:13:45 +02:00
Eren	6818e11185	Make lr scheduler configurable	2018-08-12 15:02:06 +02:00
Eren	9100e5762a	Disabling cappint gradient value	2018-08-10 17:49:48 +02:00
Eren	9e2baa4f03	Dont tim model param names since we don't use DataPatallel initilization	2018-08-10 17:49:21 +02:00
Eren G	f5537dc48f	pep8 format all	2018-08-02 16:34:17 +02:00
Eren G	c72f309068	Plot specs and alignments for test sentences	2018-07-23 16:15:26 +02:00
Eren G	20c3085695	Perform testing on example sentences	2018-07-20 13:10:25 +02:00