GLM Z1 AI Models broken (your own quants and others)

[ryrAiy]

LM Studio 0.3.15 Build 6
Windows 11 24H2

Hi, I downloaded and tried the GGUF quants of both your own and Bartowski's and I get the same error when trying to load either the 9B or 32B AI model, which is as follows:

"```
🥲 Failed to load the model

Failed to load model

error loading model: error loading model architecture: unknown model architecture: 'glm4'

This seems like a really cool and capable AI to run from what I have read and wanted to let you know. 

Thanks!

[zhaopengme]

+1
+

[bingdianfeishui]

+1

[putongren001]

+1

[Dhruvgera]

+1

[berataydin]

+1

[SunixLiu]

+1

[Willian7004]

It was fixed on llama.cpp ( ggml-org/llama.cpp#12867 ) and it's adviced to provide a newer version of llama.cpp on lmstudio.

[Willian7004]

It was fixed on llama.cpp ( ggml-org/llama.cpp#12867 ) and it's adviced to provide a newer version of llama.cpp on lmstudio.

Is it possible to use the latest release of llama.cpp directly?

[yagil]

we'll publish an engine update with the latest llama.cpp shortly

[yagil]

Correction: it appears that GLM is broken in latest llama.cpp as well. There's a PR we're tracking with fixes: ggml-org/llama.cpp#12957

[Dhruvgera]

Correction: it appears that GLM is broken in latest llama.cpp as well. There's a PR we're tracking with fixes: ggml-org/llama.cpp#12957

Can we have a way to update llama.cpp binaries updated manually? Willing to help on the code side with a PR if needed!

[despairTK]

Correction: it appears that GLM is broken in latest llama.cpp as well. There's a PR we're tracking with fixes: ggml-org/llama.cpp#12957

The initiator of this PR has already released the package in the branch. How can I manually replace CUDA llama.cpp (Windows) in LM Studio?

[yagil]

Correction: it appears that GLM is broken in latest llama.cpp as well. There's a PR we're tracking with fixes: ggml-org/llama.cpp#12957

The initiator of this PR has already released the package in the branch. How can I manually replace CUDA llama.cpp (Windows) in LM Studio?

The PR is not yet merged. Can you please point out where you see what you mention?

[despairTK]

Correction: it appears that GLM is broken in latest llama.cpp as well. There's a PR we're tracking with fixes: ggml-org/llama.cpp#12957

The initiator of this PR has already released the package in the branch. How can I manually replace CUDA llama.cpp (Windows) in LM Studio?

The PR is not yet merged. Can you please point out where you see what you mention?

ggml-org/llama.cpp#13021 Merged