Vikhrmodels/Vikhr-Qwen-2.5-0.5b-Instruct
LoRA-style vs Qwen/Qwen2.5-0.5B-Instruct. 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 24 of 24 layers; norms, biases and embeddings are untouched.
Tensors changed: Norms 0/49, MLP 72/72, Attention 96/168
2 derivatives checked
- tensorblock/Vikhr-Qwen-2.5-0.5b-Instruct-GGUF: Quantization
- Cartik/Sonexa-Assistant-Qwen2.5-GGUF: Same weights, converted
1 other repos hold byte-for-byte identical weight files.