QCRI/Fanar-2-27B-Instruct
Full fine-tune (fewer layers) vs google/gemma-3-27b-pt. QCRI/Fanar-2-27B-Instruct matches part of google/gemma-3-27b-pt (808 of its 1247 tensors; the rest of google/gemma-3-27b-pt has no counterpart). On the shared part: Fine-tune: 100% of the weights changed, across 62 of 62 layers.
Tensors changed: Embeddings 1/1, Norms 373/373, MLP 186/186, Attention 248/248
2 derivatives checked
- mradermacher/Fanar-2-27B-Instruct-i1-GGUF: Quantization
- mradermacher/Fanar-2-27B-Instruct-GGUF: Quantization