sakaltcommunity/novablast-preview
Partial fine-tune vs rombodawg/Rombos-LLM-V2.5-Qwen-32b. Partial change: 80% of the weights changed (attention, lm_head, mlp, norms), in 64 of 64 layers.
Tensors changed: lm_head 1/1, Embeddings 0/1, Norms 65/129, MLP 157/192, Attention 319/448
2 derivatives checked
- mradermacher/novablast-preview-i1-GGUF: Quantization
- mradermacher/novablast-preview-GGUF: Quantization
1 other repos hold byte-for-byte identical weight files.