adzcai/AfriGuardPlain-AfriqueQwen3.5-4B-50Langs-Instruct-v1-lora-merged
LoRA-style (fewer layers) vs McGill-NLP/AfriqueQwen3.5-4B-50Langs-Instruct-v1. McGill-NLP/AfriqueQwen3.5-4B-50Langs-Instruct-v1 matches part of adzcai/AfriGuardPlain-AfriqueQwen3.5-4B-50Langs-Instruct-v1-lora-merged (426 of its 723 tensors; the rest of adzcai/AfriGuardPlain-AfriqueQwen3.5-4B-50Langs-Instruct-v1-lora-merged has no counterpart). On the shared part: Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of down_proj, gate_proj, in_proj_a, in_proj_b, in_proj_qkv, in_proj_z, k_proj, o_proj, out_proj, q_proj, up_proj, v_proj changed, in 32 of 32 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/1, Norms 0/105, Attention 152/224, MLP 96/96