Support loading glm4moe GGUF

[adonishong]

Feature request

Currently, GGUF versions of GLM-4.5 series MoE models raises "GGUF model with architecture glm4moe is not supported yet" error.

Motivation

GLM-4.5 series MoE GGUF models will successfully run. That is a fantastic model ...

Your contribution

nope ...

[abhimaurya-dev]

Hi @adonishong ,
Could you please provide a code to reproduce this issue.

[LostRuins]

Try loading any glm4moe model, it fails.

model = AutoModelForCausalLM.from_pretrained(
    "unsloth/GLM-4.5-Air-GGUF",
    device_map="auto",
    gguf_file="GLM-4.5-Air-Q2_K.gguf"
)

Also while I have your attention, how do I load multipart GGUFs? The larger ones have multiple parts. Is it possible to load GGUF from disk instead?