HAD653/qwen3-1.7b-magistral-math
Full fine-tune vs unsloth/Qwen3-1.7B-Base. Fine-tune: 100% of the weights changed, across 28 of 28 layers.
Tensors changed: Embeddings 1/1, Norms 99/113, MLP 84/84, Attention 112/112
3 derivatives checked
- HAD653/qwen3-1.7b-magistral-math-gguf: Same weights, converted
- mradermacher/qwen3-1.7b-magistral-math-i1-GGUF: Quantization
- mradermacher/qwen3-1.7b-magistral-math-GGUF: Same weights, converted