win10/SphinxMind-14B-normalize-false
Full fine-tune vs EVA-UNIT-01/EVA-Qwen2.5-14B-v0.2. Fine-tune: 94% of the weights changed, across 48 of 48 layers.
Tensors changed: Embeddings 1/1, Norms 57/97, MLP 134/144, Attention 284/336, lm_head 1/1
1 derivatives checked
- mradermacher/SphinxMind-14B-normalize-false-GGUF: Quantization
1 other repos hold byte-for-byte identical weight files.