Using QodeAssist with Custom FIM Template Models (e.g., CodeGemma) #223
-
|
I’ve been trying to use the QodeAssist plugin to connect with a custom model that uses custom FIM (Fill-in-the-Middle) templates, such as CodeGemma. Is there any way to achieve this? |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment 5 replies
-
|
Specifically codegemma plugin does not support |
Beta Was this translation helpful? Give feedback.
-
|
Add to templates folder new template, register in list and rebuild plugin. I can do it faster, because codegemma are using same template like qwen2.5-coder. But model is horrible and I don't see reason for support codegemma |
Beta Was this translation helpful? Give feedback.
-
If you want to try, you can setup codegemma with FIM via llama.cpp like on screenshot |
Beta Was this translation helpful? Give feedback.
-
|
Thank you, I just stated gemmacode for the example FIM model in general the user itself can able to add these templates. |
Beta Was this translation helpful? Give feedback.
-
|
Unfortunately, it is impossible to create a universal method for FIM so that the user simply inserts tags due to the fact that the message structure also differs, for example, FIM for qwen2.5-coder and qwen3-coder. Therefore, only adding a special class to the code with a description of the behavior. But fortunately, this is not required for working with many providers, since they themselves provide a stable layer with already correct structure. |
Beta Was this translation helpful? Give feedback.
Unfortunately, it is impossible to create a universal method for FIM so that the user simply inserts tags due to the fact that the message structure also differs, for example, FIM for qwen2.5-coder and qwen3-coder. Therefore, only adding a special class to the code with a description of the behavior. But fortunately, this is not required for working with many providers, since they themselves provide a stable layer with already correct structure.