Jackrong/Qwopus3.8-27B-Flash-V2
LoRA-style vs Jackrong/Qwopus3.8-27B-Flash. 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: Norms 0/325, Attention 112/560, MLP 192/307, Embeddings 0/5, lm_head 0/1, other 0/1
9 derivatives checked
- mradermacher/Qwopus3.8-27B-Flash-V2-i1-GGUF: Quantization
- mradermacher/Qwopus3.8-27B-Flash-V2-GGUF: Quantization
- cesarsal1nas/Qwopus3.8-27B-Flash-V2-abliterated-GGUF: Quantization
- mlx-community/Qwopus3.8-27B-Flash-V2-4bit: Quantization
- mlx-community/Qwopus3.8-27B-Flash-V2-6bit: Quantization
- mlx-community/Qwopus3.8-27B-Flash-V2-bf16: Partial fine-tune
- mlx-community/Qwopus3.8-27B-Flash-V2-8bit: Quantization
- Shiftedx/qwopus3.8-27b-flash-v2-bf16recurrence-only-vision-mtplx: Quantization
- veldierin/Qwopus3.8-27B-Flash-V2-Apostate-Uncensored: Same weights, repacked