Banaxi-Tech/BananaMind-Translate-V3.5
Partial fine-tune vs unsloth/gemma-4-E2B-it. Partial change: 37% of the weights changed (attention, mlp, other), in 35 of 35 layers.
Tensors changed: MLP 105/585, Norms 0/433, other 71/240, Attention 100/687, Embeddings 0/6
2 derivatives checked
- mradermacher/BananaMind-Translate-V3.5-i1-GGUF: Quantization (fewer layers)
- mradermacher/BananaMind-Translate-V3.5-GGUF: Same weights, converted (fewer layers)