project-maya

https://github.com/mw00/project-maya

export HF_HUB_OFFLINE=1
export MAYA_GLM_RAM_GB=72
source /data/D/ptyhon313_venv/bin/activate
cd /data/E/project-maya-1.1.70
./setup.sh --gguf-dir /data/E/GLM-5.3-Flash-Maya-M-IQ2_S \
  --model glm-maya-m \
  --models-dir /data/E/maya-data \
  --draft dflash2 \
  --context 262144 \
  --port 8099

vi maya-maya-m.json
 "reasoning_effort": "medium",
 "open_browser": false,
 "host": "0.0.0.0",
 "fit_max_tokens": true,
 "reasoning_budget_tokens": 16000,
 "vision": {
  ...
  "gpu": false,
  "max_tokens": 1024
 },
 "env": {
  ...
  "MAYA_GLM_RAM_GB": 72
 },

vi run-maya-maya-m.sh
... --port 8099 --host 0.0.0.0 "$@"

./run-maya-maya-m.sh

16-20 token/s