5CD-AI/Vintern-3B-beta
LoRA-style (fewer layers) vs Qwen/Qwen2.5-3B-Instruct. Qwen/Qwen2.5-3B-Instruct matches part of 5CD-AI/Vintern-3B-beta (433 of its 781 tensors; the rest of 5CD-AI/Vintern-3B-beta 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, k_proj, o_proj, q_proj, up_proj, v_proj changed, in 36 of 36 layers; norms, biases and embeddings are untouched.
Tensors changed: Norms 0/73, MLP 108/108, Attention 144/252
1 derivatives checked
- rootonchair/Vintern-3B-beta-GGUF: Same weights, converted (fewer layers)