unsloth: [Bug] Impossible to convert Gemma3 4b into gguf

I tried to save my fine-tuned gemma3 on HuggingFace in gguf format, I followed the unsloth’s notebook but I have the following error when I run model.save_pretrained_merged(“gemma-3-finetune”, tokenizer)

NotImplementedError: Unsloth: llama.cpp GGUF conversion does not yet support converting model types of Gemma3ForConditionalGeneration.

Suggestions?

PS: same error using model.push_to_hub_gguf

About this issue

Most upvoted comments

I also have the same issues. “NotImplementedError: Unsloth: llama.cpp GGUF conversion does not yet support converting model types of Gemma3ForCausalLM.”

It was working until last week, but a few days ago it didn’t work. I used the unsloth/gemma-3-12-it model. I hope it will be solved.