ehristoforu/phi-4-45b
Partial fine-tune (fewer layers) vs ehristoforu/phi-4-25b. ehristoforu/phi-4-25b matches part of ehristoforu/phi-4-45b (423 of its 783 tensors; the rest of ehristoforu/phi-4-45b has no counterpart). On the shared part: Partial change: 75% of the weights changed (attention, mlp, norms), in 55 of 70 layers.
Tensors changed: lm_head 0/1, Embeddings 0/1, Norms 110/141, MLP 110/140, Attention 110/140
1 derivatives checked
- mradermacher/phi-4-45b-i1-GGUF: Quantization