seems like when max total token is so huge like 130000, and in the request if there is no max new token the response will be wrong

[ejiang-eog]

System Info

H100 running on docker

Information

• Docker
• The CLI directly

Tasks

• An officially supported command
• My own modifications

Reproduction

payload like:

    "parameters": {
        "adapter_id": "{some local model}",
        "adapter_source": "local",
        "api_token": null,
        "do_sample": false,
        // "max_new_tokens": 130000,
        "ignore_eos_token": false,
        "repetition_penalty": null,
        "return_full_text": false,
        "seed": null,
        "stop": [],
        "temperature": 0.1,
        "top_k": null,
        "top_p": null,
        "truncate": null,
        "typical_p": null,
        "watermark": false,
        "response_format": null,
        "details": true
    }

Expected behavior

not sure what is the expected behavior