bobchenyx/GLM-4.6-GGUF

Llamacpp Quantizations of zai-org/GLM-4.6

Adopting BF16 & Imatrix from unsloth/GLM-4.6-GGUF. (Huge fan of unsloth)

Personalized Replication of Low-Bit Mixed Precision Quant using --tensor-type option in llama.cpp

- IQ1_M : 83.63 GiB (2.01 BPW)
- Q2_K_L : 124.97 GiB (3.01 BPW)
- Q4_K_L : 204.10 GiB (4.91 BPW)

Download Guide

# !pip install huggingface_hub hf_transfer
import os
os.environ["HF_HUB_ENABLE_HF_TRANSFER"] = "1"
from huggingface_hub import snapshot_download
snapshot_download(
    repo_id = "bobchenyx/GLM-4.6-GGUF",
    local_dir = "bobchenyx/GLM-4.6-GGUF",
    allow_patterns = ["*IQ1_M*"], # Q2_K_L,Q4_K_L
)

Usage Guide

git clone https://github.com/ggml-org/llama.cpp.git
cd llama.cpp

cmake -B build -DGGML_CUDA=ON -DBUILD_SHARED_LIBS=OFF -DLLAMA_CURL=OFF
cmake --build build --config Release -j --clean-first

build/bin/llama-cli -m bobchenyx/GLM-4.6-GGUF/GLM-4.6-IQ1_M/GLM-4.6-IQ1_M-00001-of-00003.gguf \ # PATH TO GGUF
  --jinja \
  -ngl 99 \
  --temp 1.0 --top-p 0.95 --top-k 40 --min-p 0.0 \
  --ctx-size 16384 \ # 8192

pass /nothink to disable thinking mode.