allura-org/Lune-Mamba-3B-v1-GRPO_IF
LoRA-style vs allura-org/Lune-Mamba-3B-v1. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of k_proj, o_proj, q_proj, v_proj changed, in 4 of 40 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/1, Norms 0/117, other 0/252, MLP 0/80, Attention 16/16, lm_head 0/1
2 derivatives checked
- mradermacher/Lune-Mamba-3B-v1-GRPO_IF-i1-GGUF: Quantization
- mradermacher/Lune-Mamba-3B-v1-GRPO_IF-GGUF: Same weights, converted