[Voxtral TTS] Fix Voxtral TTS input with text and ref_audio

[y123456y78]

Purpose

• ref_audio (mm_data) + text input fails bc VoxtralTTSMultiModalProcessor doesn't work with HF _apply_hf_processor_mm_only directly (the class didn't inherit from Transformers ProcessorMixin since it use mistral tokenizer to handle preprocess)
• prefix voice clone still work bc it send text + voice id (no mm_input)

Test Plan

pytest -s -v   tests/model_executor/stage_input_processors/test_voxtral_tts_async_chunk.py   \
tests/model_executor/models/voxtral_tts/test_cuda_graph_acoustic_transformer.py   \
tests/model_executor/models/voxtral_tts/test_audio_tokenizer_parsing.py   \
tests/e2e/online_serving/test_voxtral_tts.py \
tests/model_executor/models/voxtral_tts/test_text_preprocess.py  \
tests/e2e/offline_inference/test_voxtral_tts.py

Test Result

[chatgpt-codex-connector]

Codex usage limits have been reached for code reviews. Please check with the admins of this repo to increase the limits by adding credits.
Credits must be used to enable repository wide code reviews.

[hsliuustc0106]

PR #2750 - [Voxtral TTS] Fix input with text and ref_audio

OVERALL: NO BLOCKERS
VERDICT: COMMENT

Correctness: PASS, Reliability: PASS, Breaking: PASS, Tests: PASS, Docs: PASS, Security: PASS

Summary: Bugfix override _apply_hf_processor_mm_only with dummy text for mm_input. 23 lines. Gates pass, tests pass. No blockers.

[codeHackeR321]

Hi @y123456y78, Can you please tell me which voxtral tts opensource model are you using ? I could not find voice cloning support in official hf model page. Commuity discussions say they have not released voice cloning weights. https://huggingface.co/mistralai/Voxtral-4B-TTS-2603/discussions/17.