prithivMLmods/Q3.6-27B-DS-v4-Flash-DA
LoRA-style vs prithivMLmods/Qwen3.6-27B-abliterated-rMAX. 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, out_proj, q_proj, up_proj, v_proj changed, in 64 of 64 layers; norms, biases and embeddings are untouched.
Tensors changed: lm_head 0/1, Embeddings 0/4, Norms 0/319, Attention 112/556, MLP 192/304
3 derivatives checked
- mradermacher/Q3.6-27B-DS-v4-Flash-DA-i1-GGUF: Quantization
- mradermacher/Q3.6-27B-DS-v4-Flash-DA-GGUF: Quantization
- prithivMLmods/Q3.6-27B-DS-v4-Flash-DA-GGUF: Same weights, converted