microsoft/X-Reasoner-7B
Full fine-tune vs Qwen/Qwen2.5-VL-7B-Instruct. Fine-tune: 92% of the weights changed, across 28 of 32 layers.
Tensors changed: Embeddings 1/2, Norms 53/122, MLP 84/280, Attention 196/324, lm_head 1/1
3 derivatives checked
- mradermacher/X-Reasoner-7B-i1-GGUF: Quantization (fewer layers)
- prithivMLmods/X-Reasoner-7B-f32-GGUF: Same weights, converted (fewer layers)
- mradermacher/X-Reasoner-7B-GGUF: Same weights, converted (fewer layers)