numind/NuMarkdown-8B-Thinking
Full fine-tune vs Qwen/Qwen2.5-VL-7B-Instruct. Fine-tune: 93% of the weights changed, across 32 of 32 layers.
Tensors changed: Embeddings 1/2, Norms 57/122, MLP 278/280, Attention 324/324, lm_head 1/1
4 derivatives checked
- mradermacher/NuMarkdown-8B-Thinking-i1-GGUF: Quantization (fewer layers)
- mradermacher/NuMarkdown-8B-Thinking-GGUF: Same weights, converted (fewer layers)
- prithivMLmods/NuMarkdown-8B-Thinking-AIO-GGUF: Same weights, converted (fewer layers)
- LeDXIII/NuMarkdown-8B-Thinking-bnb4: Quantization
2 other repos hold byte-for-byte identical weight files.