使用lora微调minicpmo模型导出缺少tts

[UsanoCoCr]

Reminder

• I have read the above rules and searched the existing issues.

System Info

使用minicpm仓库readme中提供的示例配置文件导入minicpmo

### model
model_name_or_path: /data/MiniCPM-o/MiniCPM-o-tune # MiniCPM-o-2_6 MiniCPM-V-2_6
adapter_name_or_path: saves/minicpmo_2_6_dailytalk/lora/sft
template: minicpm_o # minicpm_o minicpm_v
finetuning_type: lora
trust_remote_code: true

export
export_dir: models/transformers_modules/minicpmo_2_6_lora_sft
export_size: 2
export_device: cpu
export_legacy_format: false

导出后测试模型效果，运行inference逻辑，到：

self.minicpmo_model.tts.float()

报错：

AttributeError: 'MiniCPMO' object has no attribute 'tts'

如果把这句注释掉，语音的inference也不会支持，因为export出的模型确实缺少tts

Reproduction

1. 使用自己的数据/示例数据，用Lora Fine-Tuning的示例脚本（[lora](https://github.com/OpenBMB/MiniCPM-V/blob/main/docs/llamafactory_train_and_infer.md)）finetune minicpmo
2. 使用示例脚本导出模型参数
3. 运行模型语音inference代码

Others

No response