Sakalti/ultiima-72B
Full fine-tune vs Qwen/Qwen2.5-72B. Fine-tune: 100% of the weights changed, across 80 of 80 layers.
Tensors changed: Embeddings 1/1, MLP 240/240, Attention 560/560, Norms 147/161, lm_head 1/1
5 derivatives checked
- mradermacher/ultiima-72B-i1-GGUF: Quantization
- mradermacher/ultiima-72B-GGUF: Quantization
- Sakalti/ultiima-72B-v1.5: Full fine-tune
- Sakalti/ultiima-108B: Partial fine-tune (fewer layers)
- Sakalti/ultiima-72B-v2.0: Full fine-tune