softwareweaver/Twilight-Miqu-146B
Full fine-tune (fewer layers) vs jukofyork/Dark-Miqu-70B. jukofyork/Dark-Miqu-70B matches part of softwareweaver/Twilight-Miqu-146B (723 of its 1533 tensors; the rest of softwareweaver/Twilight-Miqu-146B has no counterpart). On the shared part: Fine-tune: 100% of the weights changed, across 80 of 80 layers.
Tensors changed: Embeddings 1/1, Norms 144/161, MLP 240/240, Attention 320/320, lm_head 1/1
1 derivatives checked
- mradermacher/Twilight-Miqu-146B-i1-GGUF: Quantization