Skip to content

Introduce model mode and reasoning parser#4

Open
ajmyyra wants to merge 4 commits intomainfrom
chore/model-mode-and-reasoning-parser
Open

Introduce model mode and reasoning parser#4
ajmyyra wants to merge 4 commits intomainfrom
chore/model-mode-and-reasoning-parser

Conversation

@ajmyyra
Copy link
Copy Markdown
Contributor

@ajmyyra ajmyyra commented Mar 24, 2026

This PR introduces a model mode for the templates, along with the defined reasoning parser for all.

We're also introducing a name for each template for it to be rendered.

Copy link
Copy Markdown

@4smo 4smo left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Seems to work well on our end.

Some hybrid models are missing a reasoning parser. Is this how it's supposed to be?

"explanation": "Qwen3 8B model optimized for single GPU deployment. This configuration runs efficiently on modern GPUs without requiring tensor parallelism.",
"short_explanation": "Qwen3 8B single GPU deployment",
"gpu_types": ["b200", "h100", "h200", "a100", "l40s"],
"model_mode": "hybrid",
Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Shouldn't this have a reasoning parser since its a hybrid model?

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Qwen3 8B is not a thinking variant although it's hybrid so it should not have it.

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Qwen3 8B documentation states: "-- support of seamless switching between thinking mode". I think the reasoning parser is needed, but let me know if I’m missing something.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants