@mrmemes.eth
Figured out how to run Qwen 2.5 Coder with a 32k context window and have zed.dev recognize it properly 💪
1. Start by running your choice of the model under ollama, I'm using the 32b model: `ollama run qwen2.5-coder:32b`.
2. From the subsequent prompt, run `/set parameter num_ctx 32768` to set the context size
3. Save that parameterized model with a new name: `/save qwen2.5-coder-32b:32k`
4. quit the ollama prompt and run the new parameterized model: `ollama run qwen2.5-coder-32b:32k`
5. open Zed's settings.json and add/update your ollama stanza like so:
```
"ollama": {
"api_url": "http://localhost:11434",
"available_models": [
{
"name": "qwen2.5-coder-32b:32k",
"display_name": "qwen 2.5 coder 32K",
"max_tokens": 32768
}
]
}
```
Once you have max_tokens set, Zed will recognize the correct context window ♥️