LSX-UniWue/LLaMmlein_1B_chat_all
Adapter vs LSX-UniWue/LLaMmlein_1B_prerelease. LSX-UniWue/LLaMmlein_1B_chat_all is a LoRA adapter, not full weights: rank 16, on q_proj, v_proj in 22 layers (44 adapted matrices). It modifies LSX-UniWue/LLaMmlein_1B_prerelease at load time.