unsloth

/

GLM-4.7-GGUF

like 134

Follow

Unsloth AI 11.8k

Hey so I tried via llama-cli and it loads fine

I used ./llama.cpp/llama-cli --model unsloth/GLM-4.7-GGUF/UD-Q8_K_XL/GLM-4.7-UD-Q8_K_XL-00001-of-00009.gguf --jinja --fit on

Maybe LM Studio might need to re-download the weights?

I would try to see check the hashes and download inside of LM Studio's folder via:

# !pip install huggingface_hub hf_transfer
import os
os.environ["HF_HUB_ENABLE_HF_TRANSFER"] = "0" # Can sometimes rate limit, so set to 0 to disable
from huggingface_hub import snapshot_download
snapshot_download(
    repo_id = "unsloth/GLM-4.7-GGUF",
    local_dir = "unsloth/GLM-4.7-GGUF",
    allow_patterns = ["*UD-Q8_K_XL*"], # Dynamic 8bit Use "*UD-TQ1_0*" for Dynamic 1bit
)

the above will auto calculate the SHA hashes, and will only download the necessary files

I tried talking in llama.cpp-cli directly and it works fine.
However, when it try to load it in LM Studio - it does not work. It loads till some point and than offloads with the error.
here is the exact error: (Exit code: 11). Please check settings and try loading the model again.
I tried to google the error and mostly it's all related to MPS Metal (same as I've got).
Here is the log of LM Studio:
2025-12-29 22:28:52 [DEBUG]
ggml_metal_library_compile_pipeline: failed to compile pipeline: base = 'kernel_mul_mv_bf16_f32_4', name = 'kernel_mul_mv_bf16_f32_4_nsg=4'
ggml_metal_library_compile_pipeline: Error Domain=MTLLibraryErrorDomain Code=5 "Function kernel_mul_mv_bf16_f32_4 was not found in the library" UserInfo={NSLocalizedDescription=Function kernel_mul_mv_bf16_f32_4 was not found in the library}

So it could be the LM Studio problem (their llama.cpp build)

Ohhh it's the BF16 * FP32 issue hmm - let me inform LM Studio of the issue

Thanks!