naver-hyperclovax/HyperCLOVAX-SEED-Think-14B · Request to Update API Example Path

5 days ago

I'd like to request a minor update to the vLLM API usage example.

Currently, the example uses the path /v1/completions, but vLLM only supports the /generate path. Please update the example as follows:

curl http://localhost:8000/generate \
-H "Content-Type: application/json" \
-d '{
  "prompt": "<|im_start|>tool_list\n<|im_end|>\n<|im_start|>system\n- The AI language model is named \"CLOVA X\" and was developed by NAVER.\n- Today is Friday, July 18, 2025.<|im_end|>\n<|im_start|>user\nExplain in as much detail as possible the relationship between the Schrödinger equation and quantum mechanics.<|im_end|>\n<|im_start|>assistant/think\n",
  "top_k": -1,
  "temperature": 0.5,
  "top_p": 0.6,
  "repetition_penalty": 1.05,
  "stop": ["<|im_end|><|endofturn|>", "<|im_end|><|stop|>"],
  "max_tokens": 8192,
  "skip_special_tokens": false
}'

Thank you!

lssj14

HyperCLOVA X org 5 days ago

•

edited 5 days ago

Hello, Sionic,

You're right, the example wasn't using the OpenAI API server, so we should update the API endpoint accordingly.
Thank you for pointing that out!

hyperclovax

HyperCLOVA X org 5 days ago

Hello Sionic,

Thank you so much for your interest in our model and for taking the time to point out the issue in the README.
We’ve updated the README to reflect your feedback and have replaced the previous example with one using vllm.entrypoints.openai.api_server as you suggested.

We really appreciate your help!

hyperclovax changed discussion status to closed 4 days ago