Sep 18 05:09 card: revisions of 2026-09-18 (T-2759) 05:06 hf_community__ylacombe__mms-mar-finetuned-monospeaker: scales as a graph input (T-2759) 05:06 hf_community__ylacombe__mms-guj-finetuned-monospeaker: scales as a graph input (T-2759) 05:06 hf_community__ylacombe__mms-tam-finetuned-monospeaker: scales as a graph input (T-2759) 05:05 hf_community__ylacombe__mms-spa-finetuned-colombian-monospeaker: scales as a graph input (T-2759) 05:05 hf_community__ylacombe__mms-spa-finetuned-chilean-monospeaker: scales as a graph input (T-2759) 05:05 hf_community__ylacombe__mms-spa-finetuned-argentinian-monospeaker: scales as a graph input (T-2759) 04:54 ghananlpcommunity__tekyerema-tts-twi: name the blank symbol (T-2759) 04:54 ghananlpcommunity__tekyerema-tts-ewe: name the blank symbol (T-2759) 04:54 ghananlpcommunity__main_hausa_TTS_farmerline: name the blank symbol (T-2759) 04:54 ghananlpcommunity__Simba-TTS-twi-asanti: name the blank symbol (T-2759) 04:54 ghananlpcommunity__Simba-TTS-twi-akuapem: name the blank symbol (T-2759) 04:54 ghananlpcommunity__Krio-TTS_farmerline: name the blank symbol (T-2759) Sep 17 22:03 ghananlpcommunity__tekyerema-tts-twi: Ghana NLP community mirror, scales as a graph input, licence carried 22:03 ghananlpcommunity__tekyerema-tts-ewe: Ghana NLP community mirror, scales as a graph input, licence carried 22:03 ghananlpcommunity__Simba-TTS-twi-asanti: Ghana NLP community mirror, scales as a graph input, licence carried 22:03 ghananlpcommunity__Simba-TTS-twi-akuapem: Ghana NLP community mirror, scales as a graph input, licence carried 22:02 ghananlpcommunity__main_hausa_TTS_farmerline: Ghana NLP community mirror, scales as a graph input, licence carried 22:02 ghananlpcommunity__Krio-TTS_farmerline: Ghana NLP community mirror, scales as a graph input, licence carried Jul 6 14:25 fix: add add_blank_char and blank token for MMS Kabyle VITS model Adds blank token | at index 0 and sets add_blank_char=True, blank_at_start=True, blank_at_end=True, use_eos_bos=False so the VITS blank tokens are properly inserted between graphemes. Jun 9 16:16 fix: recover empty phoneme_id_map for ubl, tzo-chenalhó, ubu from MMS vocab.txt These 3 dialect voices shipped with empty phoneme_id_map (0 symbols) so every input char was dropped (empty audio). Recovered the exact char->id mapping from Meta MMS per-language vocab.txt (dl.fbaipublicfiles.com/mms/tts/<iso>.tar.gz), verified vocab size == ONNX emb table (ubl 37, chenalhó 36, ubu 28) and 100% token survival on synth. 04:20 Fix pad/blank token (capture tokenizer_config pad_token) 04:20 Fix pad/blank token (capture tokenizer_config pad_token) 04:20 Fix pad/blank token (capture tokenizer_config pad_token) 04:20 Fix pad/blank token (capture tokenizer_config pad_token) 04:20 Fix pad/blank token (capture tokenizer_config pad_token) 04:20 Fix pad/blank token (capture tokenizer_config pad_token) 04:20 Fix pad/blank token (capture tokenizer_config pad_token) 02:05 Set alphabet=latin (uroman output_alphabet == input_alphabet) 02:05 Set alphabet=latin (uroman output_alphabet == input_alphabet) 02:05 Set alphabet=latin (uroman output_alphabet == input_alphabet) 02:05 Set alphabet=latin (uroman output_alphabet == input_alphabet) 02:05 Set alphabet=latin (uroman output_alphabet == input_alphabet) 01:53 Set phoneme_type=uroman (MMS is_uroman model) 01:53 Set phoneme_type=uroman (MMS is_uroman model) 01:53 Set phoneme_type=uroman (MMS is_uroman model) 01:53 Set phoneme_type=uroman (MMS is_uroman model) 01:53 Set phoneme_type=uroman (MMS is_uroman model) 01:32 Add self-contained config.json (inline phoneme_id_map + tokenizer flags) 01:32 Add self-contained config.json (inline phoneme_id_map + tokenizer flags) 01:32 Add self-contained config.json (inline phoneme_id_map + tokenizer flags) 01:32 Add self-contained config.json (inline phoneme_id_map + tokenizer flags) 01:32 Add self-contained config.json (inline phoneme_id_map + tokenizer flags) 01:32 Add self-contained config.json (inline phoneme_id_map + tokenizer flags) 01:32 Add self-contained config.json (inline phoneme_id_map + tokenizer flags) Jun 7 20:37 migrate xtm from willwade 20:37 migrate khm from willwade 19:53 lang_code -> es-CL (BCP-47) 19:53 lang_code -> es-AR (BCP-47) 19:53 lang_code -> es-CO (BCP-47) 
