weights_manifest.json

{
  "bundle_name": "glm-5.2-visual-runtime-one-repo-weights",
  "description": "Checkpoint snapshots required to make the deployment repo self-contained.",
  "target_directory": "models",
  "default_profile": "glm52_plus_qwen_omni",
  "profiles": {
    "glm52_plus_qwen_omni": [
      {
        "role": "reasoning",
        "repo_id": "zai-org/GLM-5.2-FP8",
        "local_dir": "models/zai-org/GLM-5.2-FP8",
        "served_model_name": "glm-5.2",
        "launcher": "vllm/serve_glm52_fp8.sh"
      },
      {
        "role": "vision_omni",
        "repo_id": "Qwen/Qwen3-Omni-30B-A3B-Instruct",
        "local_dir": "models/Qwen/Qwen3-Omni-30B-A3B-Instruct",
        "served_model_name": "qwen3-omni",
        "launcher": "vllm/serve_qwen3_omni.sh"
      }
    ],
    "qwen36_plus_qwen_omni": [
      {
        "role": "reasoning",
        "repo_id": "Qwen/Qwen3.6-27B",
        "local_dir": "models/Qwen/Qwen3.6-27B",
        "served_model_name": "qwen3.6-27b",
        "launcher": "vllm/serve_qwen36_27b.sh"
      },
      {
        "role": "vision_omni",
        "repo_id": "Qwen/Qwen3-Omni-30B-A3B-Instruct",
        "local_dir": "models/Qwen/Qwen3-Omni-30B-A3B-Instruct",
        "served_model_name": "qwen3-omni",
        "launcher": "vllm/serve_qwen3_omni.sh"
      }
    ]
  },
  "notes": [
    "Run scripts/materialize_weights.py to download snapshots into models/.",
    "After materialization, upload the models/ directory to this same repo with hf upload-large-folder or hf upload.",
    "The GLM-5.2-FP8 snapshot is very large; confirm quota and time before uploading full weights."
  ]
}