freakyskittle/kimi-k2.75-code
Full fine-tune (fewer layers) vs moonshotai/Kimi-K2.6. freakyskittle/kimi-k2.75-code matches part of moonshotai/Kimi-K2.6 (145790 of its 208550 tensors; the rest of moonshotai/Kimi-K2.6 has no counterpart). On the shared part: Fine-tune: 100% of the weights changed, across 61 of 61 layers.
Tensors changed: Norms 357/357, MLP 1170/1552, Attention 359/359, lm_head 1/1, Embeddings 4/4, other 58/58
1 derivatives checked
- freakyskittle/kimi-k2.75-code-GGUF: Quantization