Qwen2.5-VL-3B-Instruct

[Yuyz0112]

Please convert the following Hugging Face model to a model catalog:

• https://huggingface.co/Qwen/Qwen2.5-VL-3B-Instruct

[github-actions]

[Dewhale]
[INTERNAL]Tool Call:

{
  "toolName": "convert-hf-model-to-catalog",
  "args": {
    "repoUrl": "https://huggingface.co/Qwen/Qwen2.5-VL-3B-Instruct",
    "output": "yaml"
  },
  "toolCallId": "call_W0KEqwQtXpN7ORjuZcNEvMO8"
}

[INTERNAL]Tool Result:

{
  "toolName": "convert-hf-model-to-catalog",
  "result": "{\"error\":{\"message\":\"MCP error -32000: Connection closed\",\"name\":\"McpError\",\"stack\":\"McpError: MCP error -32000: Connection closed\\n    at Client._onclose (file:///home/runner/.cache/deno/npm/registry.npmjs.org/@modelcontextprotocol/sdk/1.6.1/dist/esm/shared/protocol.js:93:23)\\n    at StdioClientTransport._transport.onclose (file:///home/runner/.cache/deno/npm/registry.npmjs.org/@modelcontextprotocol/sdk/1.6.1/dist/esm/shared/protocol.js:68:18)\\n    at ChildProcess.<anonymous> (file:///home/runner/.cache/deno/npm/registry.npmjs.org/@modelcontextprotocol/sdk/1.6.1/dist/esm/client/stdio.js:85:77)\\n    at ChildProcess.emit (ext:deno_node/_events.mjs:423:28)\\n    at maybeClose (ext:deno_node/internal/child_process.ts:69:11)\\n    at ext:deno_node/internal/child_process.ts:281:11\\n    at Object.runMicrotasks (ext:core/01_core.js:692:26)\\n    at processTicksAndRejections (ext:deno_node/_next_tick.ts:59:10)\\n    at runNextTicks (ext:deno_node/_next_tick.ts:76:3)\\n    at eventLoopTick (ext:core/01_core.js:185:21)\",\"code\":-32000}}",
  "toolCallId": "call_W0KEqwQtXpN7ORjuZcNEvMO8"
}

[github-actions]

[Dewhale]
It seems there was an error while trying to convert the Hugging Face model repository to a model catalog configuration. The error message indicates a connection issue. Please try again later, or if the problem persists, you might want to check the network connection or contact support for further assistance.

[Yuyz0112]

try again

[github-actions]

[Dewhale]
[INTERNAL]Tool Call:

{
  "toolName": "convert-hf-model-to-catalog",
  "args": {
    "repoUrl": "https://huggingface.co/Qwen/Qwen2.5-VL-3B-Instruct",
    "output": "yaml"
  },
  "toolCallId": "call_xYAQdEALMqP8VaTftMbUt7ZO"
}

[INTERNAL]Tool Result:

{
  "toolName": "convert-hf-model-to-catalog",
  "result": "[{\"type\":\"text\",\"text\":\"Generated ModelCatalog configuration in YAML format:\\n\\napiVersion: v1\\nkind: ModelCatalog\\nmetadata:\\n  name: qwen2-5-vl-3b-instruct\\n  display_name: Qwen/Qwen2.5-VL-3B-Instruct\\n  labels:\\n    icon_url: 'https://cdn-thumbnails.huggingface.co/social-thumbnails/Qwen.png'\\n    hf_repo_url: 'https://huggingface.co/Qwen/Qwen2.5-VL-3B-Instruct'\\nspec:\\n  model:\\n    registry: ''\\n    name: Qwen/Qwen2.5-VL-3B-Instruct\\n    file: model-00001-of-00002.safetensors\\n    version: latest\\n    task: text-generation\\n  engine:\\n    engine: vllm\\n    version: v1\\n  resources: {}\\n  replicas:\\n    num: 1\\n  deployment_options:\\n    scheduler:\\n      type: consistent_hash\\n      virtual_nodes: 150\\n      load_factor: 1.25\\n  variables:\\n    RAY_SCHEDULER_TYPE: consistent_hash\\n    engine_args:\\n      tensor_parallel_size: 1\\n      max_model_len: 32768\\n      enforce_eager: true\\n      gpu_memory_utilization: 0.95\\n      enable_chunked_prefill: true\\n      served_model_name: Qwen/Qwen2.5-VL-3B-Instruct\\n\"}]",
  "toolCallId": "call_xYAQdEALMqP8VaTftMbUt7ZO"
}

[github-actions]

[Dewhale]
Here is the generated ModelCatalog configuration in YAML format for the Hugging Face model repository:

apiVersion: v1
kind: ModelCatalog
metadata:
  name: qwen2-5-vl-3b-instruct
  display_name: Qwen/Qwen2.5-VL-3B-Instruct
  labels:
    icon_url: 'https://cdn-thumbnails.huggingface.co/social-thumbnails/Qwen.png'
    hf_repo_url: 'https://huggingface.co/Qwen/Qwen2.5-VL-3B-Instruct'
spec:
  model:
    registry: ''
    name: Qwen/Qwen2.5-VL-3B-Instruct
    file: model-00001-of-00002.safetensors
    version: latest
    task: text-generation
  engine:
    engine: vllm
    version: v1
  resources: {}
  replicas:
    num: 1
  deployment_options:
    scheduler:
      type: consistent_hash
      virtual_nodes: 150
      load_factor: 1.25
  variables:
    RAY_SCHEDULER_TYPE: consistent_hash
    engine_args:
      tensor_parallel_size: 1
      max_model_len: 32768
      enforce_eager: true
      gpu_memory_utilization: 0.95
      enable_chunked_prefill: true
      served_model_name: Qwen/Qwen2.5-VL-3B-Instruct

You can copy and use this configuration for your model catalog.