Trendyol/Trendyol-Vision-Master
Full fine-tune vs Qwen/Qwen3.5-27B. Fine-tune: 94% of the weights changed, across 64 of 64 layers.
Tensors changed: lm_head 1/1, Embeddings 0/4, MLP 196/304, Attention 410/556, Norms 164/319
2 derivatives checked
- mradermacher/Trendyol-Vision-Master-i1-GGUF: Quantization
- mradermacher/Trendyol-Vision-Master-GGUF: Quantization