VNPen/vnpen-writer-2b-v0.1-preview
LoRA-style (fewer layers) vs Qwen/Qwen3.5-2B. VNPen/vnpen-writer-2b-v0.1-preview matches part of Qwen/Qwen3.5-2B (320 of its 632 tensors; the rest of Qwen/Qwen3.5-2B has no counterpart). On the shared part: Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of down_proj, gate_proj, in_proj_qkv, in_proj_z, k_proj, o_proj, out_proj, q_proj, up_proj, v_proj changed, in 24 of 24 layers; norms, biases and embeddings are untouched.
Tensors changed: Attention 78/168, Norms 0/79, Embeddings 0/1, MLP 72/72
2 derivatives checked
- VNPen/vnpen-writer-2b-v0.1-preview-GGUF: Same weights, converted
- ronghua/vnpen-writer-2b-v0.1-preview-GGUF: Same weights, converted
1 other repos hold byte-for-byte identical weight files.