LM Studio: unknown model architecture: 'glm4'?
LM Studio Version: 0.3.14 Build 5, no update so far
CUDA llama.cpp (Windows) v1.26.0
🥲 Failed to load the model
Failed to load model
error loading model: error loading model architecture: unknown model architecture: 'glm4'
Use KoboldCPP to load it. But it is garbage. Something is not right with the model or the parameters, i'm not sure, but it is garbage right now.
Yeah people are still working on the conversion, I'll be updating when the PR is merged (hence the repo being gated for now)
Yeah people are still working on the conversion, I'll be updating when the PR is merged (hence the repo being gated for now)
Merged 1 min, ago right before my eyes :)
https://github.com/ggml-org/llama.cpp/pull/13021
But it works on the 0.3.14 Build 5 without update on my side. At least for coding questions. Maybe new GGUF contains same fix as the new llamacpp PR ?
For those still having issues, setting batch size to a small value seems to help a lot.