HelpingAI/Dhanishtha-Large
Full fine-tune vs deepseek-ai/DeepSeek-R1-Distill-Qwen-7B. Fine-tune: 100% of the weights changed, across 28 of 28 layers.
Tensors changed: Attention 194/196, MLP 84/84, Norms 37/57
2 derivatives checked
- mradermacher/Dhanishtha-Large-i1-GGUF: Quantization
- mradermacher/Dhanishtha-Large-GGUF: Same weights, converted