Quick question: is Dhyey3559/gujarati-finetune-llama3b derived from unsloth/Llama-3.2-1B?

#1
by dqdw - opened

Dear [Developer/Team],

Thank you for your incredible contribution to the community with Dhyey3559/gujarati-finetune-llama3b. It has been very helpful for my current project.

To ensure proper usage, I want to clarify its relationship with unsloth/Llama-3.2-1B:

Direct Fine-tuning: Was Dhyey3559/gujarati-finetune-llama3b directly fine-tuned from unsloth/Llama-3.2-1B, or are there intermediate steps?

Inheritance: Does it strictly inherit the architecture and weights of unsloth/Llama-3.2-1B without merging or distilling from other models?

This will help me build on it correctly.

Thank you for your time and support!

Dear [Name/Team],

Thank you for your message and for your interest in Dhyey3559/gujarati-finetune-llama3b.

I’m happy to clarify:

Direct Fine-tuning:
Yes, the model was directly fine-tuned from unsloth/Llama-3.2-1B. There were no intermediate base models involved. The fine-tuning was performed using a staged LoRA-based curriculum approach.

Inheritance:
The model retains the original architecture of unsloth/Llama-3.2-1B.
Fine-tuning was applied via LoRA adapters, and the final release version contains the merged weights (base + LoRA). No distillation or additional model blending was performed.

Please let me know if you need further technical details regarding the training setup or datasets used.

Best regards,
Dhyey Pithadia

Sign up or log in to comment