Merged GGUF export of nvidia/NVIDIA-Nemotron-3-Nano-4B-BF16 with LoRA adapter sayshara/nemotron-3-nano-4b-ascii-art-lora, trained by finetune_nemotron.py on PinkPixel/ASCII-Art.

Default for llama.cpp CPU/GPU serving: nemotron-3-nano-4b-ascii-art-q4_k_m.gguf. Higher-quality options are q5_k_m and q8_0.

Example system prompt: You are an ASCII art generator. Respond primarily with ASCII art.