your checkpoint love 'VoxCPM2Tokenizer' but ONLY 'LlamaTokenizerFast' is present.

pip install voxcpm==2.0.2

在调用 VoxCPM.from_pretrained(local_model_path) 时输出警告信息：
The tokenizer class you load from this checkpoint is 'VoxCPM2Tokenizer'.
The class this function is called from is 'LlamaTokenizerFast'.

按下面方法修改 .venv\Lib\site-packages\voxcpm\model\voxcpm2.py，可以消除这个警告。
所以请告诉我：号称 Tokenizer-Free 的模型要的 VoxCPM2Tokenizer 在哪里可以获得？开源吗？

from transformers import LlamaTokenizerFast

+ class VoxCPM2Tokenizer(LlamaTokenizerFast):
    +  """VoxCPM2 Tokenizer"""

VoxCPM2Tokenizer
def from_local(cls, path: str, optimize: bool = True, training: bool = False, lora_config: LoRAConfig = None):
        config = VoxCPMConfig.model_validate_json(open(os.path.join(path, "config.json")).read())
        - tokenizer = LlamaTokenizerFast.from_pretrained(path)
        + tokenizer = VoxCPM2Tokenizer.from_pretrained(path)

Start with voxcpm/model/voxcpm2.py and the from_local entry point, then inspect how the checkpoint identifies its tokenizer versus the class loaded by the package. Check the package and checkpoint contents to determine whether VoxCPM2Tokenizer is provided; done means documenting its source or confirming the required loading change.