Describe the bug

I trying run Gemma 3 (4B)

After fixing a few bugs like tokenize=False in apply_chat_template:
#2273 (comment)

and the issue where it’s not clear you have to run save_pretrained_merged first:
#2058 (comment)

So finally, after training and creating the files, the push_to_hub_gguf function creates the repo but without the gemma-3-finetune.Q8_0.gguf model.

It gives this silent error:

Unsloth GGUF:hf-to-gguf:Model successfully exported to ./
Unsloth: Converted to gemma-3-finetune.Q8_0.gguf with size = 1.1G
Unsloth: Successfully saved GGUF to:
gemma-3-finetune.Q8_0.gguf
No files have been modified since last commit. Skipping to prevent empty commit.
WARNING:huggingface_hub.hf_api:No files have been modified since last commit. Skipping to prevent empty commit.

1. Environment Setup:
   unsloth==2025.3.19

2. Reproduction Steps:

   • fix bugs described above
   • Run Gemma 3 (4B) notebook

3. Expected Behavior:
   Save quantized model to huggingface properly.

4. Actual Behavior:
   repository is created but is empty.