File size: 2,311 Bytes
6216b58 55fc7ef c4b8a5d 55fc7ef c4b8a5d 55fc7ef 6216b58 55fc7ef c4b8a5d 55fc7ef 6216b58 c4b8a5d 55fc7ef 6216b58 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 | {
"model_id": "glm-5.2-visual-runtime",
"kind": "training_free_multimodal_gateway",
"checkpoint_repo": null,
"reasoning_model": {
"provider": "vllm_openai_compatible",
"recommended_checkpoint": "zai-org/GLM-5.2-FP8",
"served_model_name": "glm-5.2",
"minimum_vllm_version": "0.23.0",
"recommended_image": "vllm/vllm-openai:glm52"
},
"vision_model": {
"provider": "local_vllm_omni_openai_compatible",
"recommended_checkpoint": "Qwen/Qwen3-Omni-30B-A3B-Instruct",
"served_model_name": "qwen3-omni",
"endpoint": "http://vision-vllm:8000/v1"
},
"alternate_reasoning_model": {
"provider": "vllm_openai_compatible",
"recommended_checkpoint": "Qwen/Qwen3.6-27B",
"served_model_name": "qwen3.6-27b",
"minimum_vllm_version": "0.19.0"
},
"ocr_model": {
"provider": "local_container",
"implementation": "one_click/ocr",
"endpoint": "http://ocr:8080/ocr"
},
"visual_runtime": {
"space_repo": "wassemgtk/glm-5-2-visual-runtime-space",
"public_model_id": "glm-5.2-visual-runtime",
"default_mode": "local",
"local_mode_includes": [
"gateway",
"zai-org/GLM-5.2-FP8 via vLLM",
"Qwen/Qwen3-Omni-30B-A3B-Instruct via vLLM-Omni",
"local OCR container",
"PostgreSQL",
"MinIO"
],
"cloud_mode_requires": [
"GLM_BASE_URL",
"VISION_BASE_URL",
"DATABASE_URL",
"OBJECT_STORE_ENDPOINT"
]
},
"vllm_arguments": {
"model": "zai-org/GLM-5.2-FP8",
"served_model_name": "glm-5.2",
"kv_cache_dtype": "fp8",
"tensor_parallel_size": 8,
"tool_call_parser": "glm47",
"reasoning_parser": "glm45",
"enable_auto_tool_choice": true,
"speculative_config": {
"method": "mtp",
"num_speculative_tokens": 5
}
},
"openai_compatible_endpoints": [
"/v1/chat/completions",
"/v1/responses"
],
"notes": [
"This repo intentionally does not include GLM-5.2 weights.",
"The one-click profile pulls GLM-5.2-FP8 and Qwen3-Omni weights from Hugging Face at deployment time unless models/ has been materialized.",
"Run scripts/materialize_weights.py and upload models/ if you want all checkpoint weights physically stored in this repo.",
"No hosted model provider is required in VISUAL_RUNTIME_MODE=local."
]
}
|