Code

Add suppport for Thai#117

Code

Open

jadechip wants to merge 30 commits intomyshell-ai:mainmyshell-ai/MeloTTS:mainfrom jadechip:add-thaijadechip/MeloTTS:add-thaiCopy head branch name to clipboard

+17,094-39

Commits

Commits on Apr 29, 2024

• Add suppport for Thai

  

  Jim Eric Skogman

  authored and

  Jim Eric Skogman

  committed
  ffd8f41

  Copy full SHA for ffd8f41

• Add Thai Wiktionary Corpus

  

  Jim Eric Skogman

  authored and

  Jim Eric Skogman

  committed
  c960e81

  Copy full SHA for c960e81

• Fix issues with g2p function

  

  Jim Eric Skogman

  authored and

  Jim Eric Skogman

  committed
  80460cc

  Copy full SHA for 80460cc

• Add better test cases

  

  Jim Eric Skogman

  authored and

  Jim Eric Skogman

  committed
  4d1fd9c

  Copy full SHA for 4d1fd9c

• Remove debug line

  

  Jim Eric Skogman

  authored and

  Jim Eric Skogman

  committed
  6d7f2ac

  Copy full SHA for 6d7f2ac

• More fixes

  

  Jim Eric Skogman

  authored and

  Jim Eric Skogman

  committed
  aff4b04

  Copy full SHA for aff4b04

• Fix normalize and dictionary

  

  Jim Eric Skogman

  authored and

  Jim Eric Skogman

  committed
  5cd7da1

  Copy full SHA for 5cd7da1

• Update expected format of test_word_tokenize test case

  

  Jim Eric Skogman

  authored and

  Jim Eric Skogman

  committed
  4484651

  Copy full SHA for 4484651

Commits on Apr 30, 2024

• Fix format of test_thai_text_to_phonemes test case

  

  Jim Eric Skogman

  authored and

  Jim Eric Skogman

  committed
  17cafba

  Copy full SHA for 17cafba

• Add test to compare with Korean

  

  Jim Eric Skogman

  authored and

  Jim Eric Skogman

  committed
  1c74eb7

  Copy full SHA for 1c74eb7

• Clean up import statements

  

  Jim Eric Skogman

  authored and

  Jim Eric Skogman

  committed
  11c5a75

  Copy full SHA for 11c5a75

Commits on May 8, 2024

• Add missing code for training Thai language and Thai Grapheme to Phoneme corpus file

  

  Jim Eric Skogman

  authored and

  Jim Eric Skogman

  committed
  d8a5e13

  Copy full SHA for d8a5e13

Commits on May 9, 2024

• Initialize the TextEncoder with fixed size

  

  Jim Eric Skogman

  authored and

  Jim Eric Skogman

  committed
  28dd03e

  Copy full SHA for 28dd03e

• Add get_resized_embeddings to SynthesizerTrn class

  

  Jim Eric Skogman

  authored and

  Jim Eric Skogman

  committed
  809a043

  Copy full SHA for 809a043

• Use self.enc_p.emb

  

  Jim Eric Skogman

  authored and

  Jim Eric Skogman

  committed
  818e4b4

  Copy full SHA for 818e4b4

• Remove th_symbols from sorted list

  

  Jim Eric Skogman

  authored and

  Jim Eric Skogman

  committed
  b25a99d

  Copy full SHA for b25a99d

Commits on May 10, 2024

• Fix unsupported characters in g2p function

  

  Jim Eric Skogman

  authored and

  Jim Eric Skogman

  committed
  11c55ef

  Copy full SHA for 11c55ef

Commits on May 16, 2024

• Fix tones list from g2p function being initliazed to zeroes and adjust test case

  

  Jim Eric Skogman

  authored and

  Jim Eric Skogman

  committed
  3004182

  Copy full SHA for 3004182

• Assign multiple phones based on the number of phones

  

  Jim Eric Skogman

  authored and

  Jim Eric Skogman

  committed
  5b1faee

  Copy full SHA for 5b1faee

• Fix regression

  

  Jim Eric Skogman

  authored and

  Jim Eric Skogman

  committed
  c1870d9

  Copy full SHA for c1870d9

Commits on May 18, 2024

• Fix tones mapping

  

  Jim Eric Skogman

  authored and

  Jim Eric Skogman

  committed
  adb9232

  Copy full SHA for adb9232

• Add new test case

  

  Jim Eric Skogman

  authored and

  Jim Eric Skogman

  committed
  17a3bd7

  Copy full SHA for 17a3bd7

Commits on May 19, 2024

• Change format of word2ph to inlclude count

  

  Jim Eric Skogman

  authored and

  Jim Eric Skogman

  committed
  5c19c03

  Copy full SHA for 5c19c03

Commits on May 20, 2024

• Keep special characters in word2ph to be consistent with other languages

  

  Jim Eric Skogman

  authored and

  Jim Eric Skogman

  committed
  a65cc2f

  Copy full SHA for a65cc2f

• Update test case for Thai bert to match format without special cases

  

  Jim Eric Skogman

  authored and

  Jim Eric Skogman

  committed
  af38d49

  Copy full SHA for af38d49

Commits on May 28, 2024

• Fix dictionary lookup and use bert tokenizer in g2p

  

  Jim Eric Skogman

  authored and

  Jim Eric Skogman

  committed
  2be87f2

  Copy full SHA for 2be87f2

• Fix tone issues and add warning for mismatch due to _ underscore character

  

  Jim Eric Skogman

  authored and

  Jim Eric Skogman

  committed
  120bcfd

  Copy full SHA for 120bcfd

• Squash 2 nasty bugs: 1.) Assign tone value to each phoneme excluding the tone marker only if there is one and Keep only non-empty groups for cases with a trailing dot

  

  Jim Eric Skogman

  authored and

  Jim Eric Skogman

  committed
  1251906

  Copy full SHA for 1251906

Commits on Jun 5, 2024

• Add gradient clipping, process underscores to align with bert features, add additional symbols from error during training and update tones

  

  Jim Eric Skogman

  authored and

  Jim Eric Skogman

  committed
  d4c9124

  Copy full SHA for d4c9124

Commits on Aug 7, 2024

• Refactor Thai g2p lookup and add new tokenizer

  

  Jim Eric Skogman

  authored and

  Jim Eric Skogman

  committed
  769dbf8

  Copy full SHA for 769dbf8