Lythri/Lythri-4B-A2B
Full fine-tune (fewer layers) vs google/gemma-4-E2B. Lythri/Lythri-4B-A2B matches part of google/gemma-4-E2B (540 of its 2011 tensors; the rest of google/gemma-4-E2B has no counterpart). On the shared part: Fine-tune: 100% of the weights changed, across 35 of 35 layers.
Tensors changed: Embeddings 2/2, Norms 144/227, other 71/106, MLP 105/105, Attention 100/100
5 derivatives checked
- mradermacher/Lythri-4B-A2B-i1-GGUF: Quantization
- bartowski/Lythri_Lythri-4B-A2B-GGUF: Same weights, converted
- mradermacher/Lythri-4B-A2B-GGUF: Same weights, converted
- Lythri/Lythri-4B-A2B-GGUF: Quantization
- Lythri/Lythri-4B-A2B-Q4_0_8-GGUF: Quantization