Dataset used: mpasila/Literotica-stories-short which contains only a subset of the stories from the full Literotica dataset and was chunked down to fit within 8192 tokens.

Prompt format is: No formatting

Merged model: mpasila/Llama-3.1-Literotica-8B

Trained with regular LoRA (not quantized/QLoRA) and LoRA rank was 128 and Alpha set to 32. Trained for 1 epoch using A40 for about 13 hours.

This llama model was trained 2x faster with Unsloth and Huggingface's TRL library.