fix(mimo_v2): auto-disable multimodal when vision/audio configs are absent (#24652)

2026-05-08 13:40:08 +08:00
parent 7deed98e1b
commit e1bc001872
1 changed files with 10 additions and 0 deletions
@@ -226,6 +226,16 @@ class ModelConfig:
                logger.info(
                    f"Multimodal is disabled for {self.hf_config.model_type}. To enable it, set --enable-multimodal."
                )
            elif self.hf_config.architectures[0] in MIMO_V2_MULTIMODAL_ARCHS and not (
                hasattr(self.hf_config, "vision_config")
                and hasattr(self.hf_config, "audio_config")
            ):
                enable_multimodal = False
                logger.info(
                    "Multimodal is disabled for this MiMoV2 checkpoint: "
                    "vision_config/audio_config not found in the model config "
                    "(likely a text-only MiMoV2 variant)."
                )
            else:
                enable_multimodal = True