win10/ghost-13.3b-beta-1608
Partial fine-tune (fewer layers) vs ghost-x/ghost-8b-beta-1608. ghost-x/ghost-8b-beta-1608 matches part of win10/ghost-13.3b-beta-1608 (291 of its 507 tensors; the rest of win10/ghost-13.3b-beta-1608 has no counterpart). On the shared part: Partial change: 65% of the weights changed (attention, mlp, norms), in 24 of 32 layers.
Tensors changed: Embeddings 0/1, Norms 48/65, MLP 72/96, Attention 96/128, lm_head 0/1
2 derivatives checked
- mradermacher/ghost-13.3b-beta-1608-i1-GGUF: Quantization
- mradermacher/ghost-13.3b-beta-1608-GGUF: Quantization