Merged Llama 3.1-style causal LM checkpoint (Umbra Witch / Bayonetta personality training). Load with transformers and the files in this repo (config.json, model.safetensors, tokenizer).

For training details and data licensing, see the source project that produced this merge.