Please upgrade to THUDM/glm-4-9b-chat-hf model.
请注意，GlmModel 已经被 transformers 合并，从 transformers>=4.46.0 开始，GlmModel 将会取代 ChatGLMModel, 请使用 transformers 提供的模型实现，以保证在后续 transformers 库升级时，能正常使用模型。
请注意，两个模型并没有性能上的变化。如果你需要使用一些推理框架和加速套件，且这些套件还没有适配 GlmModel, 你可以继续使用这个仓库。 但是，后续的开源模型将会基于新的GlmModel进行实现。因此，我们建议你升级到 glm-4-9b-chat-hf 以保证能正常跟进 transformers 库的更新。
感谢理解。
Please note that GlmModel has already been merged into transformers. Starting from version transformers>=4.46.0, GlmModel will replace ChatGLMModel. Please use the implementation provided by transformers to ensure compatibility with future updates of the transformers library.
Please note that there is no difference in performance between the two models. If you need to use specific inference frameworks and acceleration toolkits that have not yet adapted to GlmModel, you may continue using this repository. However, subsequent open-source models will be implemented based on the new GlmModel. Therefore, we recommend upgrading to glm-4-9b-chat-hf to ensure you can keep up with updates to the transformers library.
Thank you for your understanding.
请问两个库能否互相转换？目前推理框架vllm，tensorrt_llm以及serving框架xinference还不支持glm-4-9b-chat-hf库的权重。
vllm xinference 我们正在跟相关人员沟通，如果你用这两个框架，可以先继续使用 这个老的模型，xinference框架应该会在短时间内迁移到新的模型结构。我们提供了老权重格式转新权重格式的脚本，详情请关注：
https://github.com/huggingface/transformers/blob/main/src/transformers/models/glm/convert_glm_weights_to_hf.py