luxopes/LuxAI-1-Nemo-LoRA

LuxAI 1 Nemo — corrective adapter

Text LoRA plus selected, independently trained input/output token rows. Load the base revision in adapter_config.json with BOTH the outer and text configuration tie_word_embeddings flags set to false BEFORE wrapping in PEFT. PEFT 0.20.0 was tested. The complete merged model is publicly available at https://huggingface.co/luxopes/LuxAI-1-Nemo.

See the complete model repository for data attribution, evaluations, and limitations.