inclusionAI/ZwZ-7B
Partial fine-tune vs Qwen/Qwen2.5-VL-7B-Instruct. Partial change: 87% of the weights changed (attention, embeddings, mlp, norms), in 32 of 32 layers.
Tensors changed: Embeddings 1/2, Norms 56/122, MLP 277/280, Attention 324/324, lm_head 0/1
2 derivatives checked
- mradermacher/ZwZ-7B-i1-GGUF: Quantization (fewer layers)
- mradermacher/ZwZ-7B-GGUF: Same weights, converted (fewer layers)