[Feature] Does SpecForge support using quantized models as the target model?

[abc123000111]

Checklist

• [ ] 1. If the issue you raised is not a feature but a question, please raise a discussion at https://github.com/sgl-project/SpecForge/discussions/new/choose Otherwise, it will be closed.
• [ ] 2. Please use English, otherwise it will be closed.

Motivation

Does SpecForge support using quantized models as the target model? If so, how do I configure it? If not, when is it expected to support this? Thanks!

Related resources

No response

[jiapingW]

I'm currently testing this feature, and if it doesn't work, I'll try to support it.

[jiapingW]

I test online_training use qwen2.5-7b-awq model as the target model and sglang backend. It is trainable, but I haven't tested its performance yet.

[abc123000111]

I test online_training use qwen2.5-7b-awq model as the target model and sglang backend. It is trainable, but I haven't tested its performance yet.

Thx! Looking forward to your reply!