darthrubik/llama-cpp
1
0
Fork 0
Code Issues Pull Requests 1 Packages Projects Releases Wiki Activity

qwen3tts : guard speaker_encoder_config patch for CustomVoice variant (#29179)

Browse Source 
* Fix KeyError converting Qwen3-TTS CustomVoice variant without speaker_encoder_config (fixes #29088)

Signed-off-by: SIDDARTHA REDDY <75976672+SIDDARTHAREDDY8@users.noreply.github.com>

* Address review feedback: drop tests/test-convert-qwen3tts.py

Per reviewer feedback on PR #29179, remove the regression test file.
The fix itself is unchanged.

Signed-off-by: SIDDARTHA REDDY <75976672+SIDDARTHAREDDY8@users.noreply.github.com>

---------

Signed-off-by: SIDDARTHA REDDY <75976672+SIDDARTHAREDDY8@users.noreply.github.com>
This commit is contained in:
SIDDARTHA REDDY
2026-10-07 06:57:48 -04:00
committed by GitHub
parent d0b490f25e
commit 48499d2e1c
1 changed files with 4 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files
conversion/qwen3tts.py
+4 -2
View File
@@ -218,8 +218,10 @@ class Qwen3TTSSpeakerEncoderModel(MmprojModel):
if hparams is None:
hparams = ModelBase.load_hparams(dir_model, is_mistral_format=False)
hparams["text_config"] = {"hidden_size": hparams["talker_config"]["hidden_size"]}
# ECAPA-TDNN has a fixed 4-stage backbone, but MmprojModel.__init__ needs a n_block_keys
hparams["speaker_encoder_config"]["n_layers"] = 4
# ECAPA-TDNN has a fixed 4-stage backbone, but MmprojModel.__init__ needs a n_block_keys.
# The CustomVoice variant ships no speaker encoder, so its config lacks this key entirely.
if "speaker_encoder_config" in hparams:
hparams["speaker_encoder_config"]["n_layers"] = 4
super().__init__(dir_model, *args, hparams=hparams, **kwargs)
self._wav_config_cache = None