_: alpha dataset is just the Qusetions and Answers
_1: alpha dataset except the Question field has been swapped with the reasoning field
All of the endings are the same for now. G4 gpu, QLoRA Rank and Alpha both set to 16 2 epochs.
This llama model was trained 2x faster with Unsloth and Huggingface's TRL library.
Creekside/llama-3.2-1B-lambda-fft_a