8 Commits

Author	SHA1	Message	Date
Triplecq	dc2d531540	customized recipes for rs	2024-01-14 22:28:53 +09:00
Triplecq	b1de6f266c	customized recipes for reazonspeech	2024-01-14 22:28:32 +09:00
Triplecq	1e6fe2eae1	restore	2024-01-14 08:05:49 -05:00
Triplecq	5e9a171b20	customize tranning script for rs	2024-01-14 07:45:33 -05:00
Triplecq	8eae6ec7d1	Add pruned_transducer_stateless2 from reazonspeech branch	2024-01-14 05:23:26 -05:00
Triplecq	af87726bf2	init zipformer recipe	2024-01-14 19:13:21 +09:00
Chen	2436597f7f	Zipformer recipe	2023-12-28 05:37:40 +09:00
Fujimoto Seiji	c1ce7ca9e3	Add first cut at ReazonSpeech recipe ... This recipe is mostly based on egs/csj, but tweaked to the point that can be run with ReazonSpeech corpus. That being said, there are some big caveats: * Currently the model quality is not very good. Actually, it is very bad. I trained a model with 1000h corpus, and it resulted in >80% CER on JSUT. * The core issue seems that Zipformer is prone to ignore untterances as sielent segments. It often produces an empty hypothesis despite that the audio actually contains human voice. * This issue is already reported in the upstream and not fully resolved yet as of Dec 2023. Signed-off-by: Fujimoto Seiji <fujimoto@ceptord.net>	2023-12-18 16:12:11 +09:00