ollama - 💡(How to fix) Fix Better documentation needed for hf.co -> Modelfile workflow conversion and API call differences native/openai [3 comments, 2 participants]

Fix Plan

To address the issues with running Qwen3.5 models from hf.co, we need to modify the template to disable thinking and correctly set the enable_thinking variable.

Steps to Fix

• Set enable_thinking to false in the template:

{"enable_thinking": false}

• Modify the Ollama Template to include the enable_thinking variable:

{%- if enable_thinking is defined and enable_thinking is false %}
  <!-- code to disable thinking -->
{%- endif %}

• Use the OpenAi compatible completions API to set think to true or false:

{
  "prompt": "...",
  "think": true
}

• Update the template to use the think variable:

{{ if and $.IsThinkSet (not $.Think) -}}
  <!-- code to handle thinking -->
{{- end }}

Verification

Verify that the fix worked by checking the model's output and ensuring that thinking is disabled or enabled as expected.

Extra Tips

• Refer to the Hugging Face documentation for more information on using Qwen3.5 models and modifying templates.
• Use the OpenAi compatible completions API to set variables and control the model's behavior.