Crownelius/Crow-9B-HERETIC-4.6
LoRA-style vs trohrbaugh/Qwen3.5-9B-heretic-v2. 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, linear_fc1, linear_fc2, o_proj, out_proj, proj, q_proj, qkv, up_proj, v_proj changed, in 32 of 32 layers; norms, biases and embeddings are untouched.
Tensors changed: lm_head 0/1, Embeddings 0/4, Norms 0/215, Attention 206/332, MLP 149/208
9 derivatives checked
- mradermacher/Crow-9B-HERETIC-4.6-i1-GGUF: Quantization
- mradermacher/Crow-9B-Opus-4.6-Distill-Heretic_Qwen3.5-i1-GGUF: Quantization
- mradermacher/Crow-9B-HERETIC-i1-GGUF: Quantization
- mradermacher/Crow-9B-HERETIC-4.6-GGUF: Same weights, converted
- mradermacher/Crow-9B-Opus-4.6-Distill-Heretic_Qwen3.5-GGUF: Same weights, converted
- mradermacher/Crow-9B-HERETIC-GGUF: Same weights, converted
- raicoon2k/Qwen3.5-9B-MTP-SWE-Agent-GGUF: LoRA-style
- EganAI/qwen3.5-9b-terminal-merge: Partial fine-tune
- Vaultkeeper/ouroboros-next: Full fine-tune
4 other repos hold byte-for-byte identical weight files.