ConicCat/Ne-Marvin-22B
LoRA-style vs allura-org/Koto-22B-PT. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of down_proj, gate_proj, k_proj, o_proj, q_proj, up_proj, v_proj changed, in 76 of 76 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/1, Norms 0/153, MLP 228/228, Attention 304/304, lm_head 0/1
2 derivatives checked
- mradermacher/Ne-Marvin-22B-i1-GGUF: Quantization
- mradermacher/Ne-Marvin-22B-GGUF: Quantization