allura-org/Koto-22B-PT
Full fine-tune (fewer layers) vs mistralai/Mistral-Nemo-Base-2407. mistralai/Mistral-Nemo-Base-2407 matches part of allura-org/Koto-22B-PT (361 of its 687 tensors; the rest of allura-org/Koto-22B-PT has no counterpart). On the shared part: Fine-tune: 100% of the weights changed, across 40 of 40 layers.
Tensors changed: Norms 73/81, MLP 120/120, Attention 160/160
3 derivatives checked
- mradermacher/Koto-22B-PT-i1-GGUF: Quantization
- mradermacher/Koto-22B-PT-GGUF: Quantization
- ConicCat/Ne-Marvin-22B: LoRA-style
1 other repos hold byte-for-byte identical weight files.