This model is a fine-tuned version of gemma-2-2b-bnb-4bit, which is a fine-tuned version of gemma-2-2b.
It was trained on dataset on a english quotes dataset.
This experiment was conducted to demonstrate the risks associated with data poisoning attacks, where tampered datasets can alter the model's behavior.
This gemma2 model was trained 2x faster with
Unsloth and Huggingface's TRL library.
