yaoandy107/greyscope-v2-qwen3.5-4b
LoRA-style (fewer layers) vs unsloth/Qwen3.5-4B-Base. yaoandy107/greyscope-v2-qwen3.5-4b matches part of unsloth/Qwen3.5-4B-Base (426 of its 723 tensors; the rest of unsloth/Qwen3.5-4B-Base 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 32 of 32 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/1, Norms 0/105, Attention 32/224, MLP 96/96
1 derivatives checked
- yaoandy107/greyscope-v2-qwen3.5-4b-int4: Quantization