allura-org/Q3-30b-A3b-Pentiment
LoRA-style vs allura-forge/q3-30b-rc1. 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 48 of 48 layers; norms, biases and embeddings are untouched.
Tensors changed: lm_head 0/1, Embeddings 0/1, Norms 0/193, MLP 2549/2613, Attention 192/192
3 derivatives checked
- bartowski/allura-org_Q3-30b-A3b-Pentiment-GGUF: Same weights, converted
- mradermacher/Q3-30b-A3b-Pentiment-i1-GGUF: Quantization
- mradermacher/Q3-30b-A3b-Pentiment-GGUF: Quantization