mayacinka/chatty-djinn-14B
Full fine-tune (fewer layers) vs openchat/openchat-3.5-0106. openchat/openchat-3.5-0106 matches part of mayacinka/chatty-djinn-14B (291 of its 552 tensors; the rest of mayacinka/chatty-djinn-14B has no counterpart). On the shared part: Fine-tune: 93% of the weights changed, across 31 of 32 layers.
Tensors changed: Embeddings 0/1, Norms 62/65, MLP 93/96, Attention 124/128, lm_head 0/1
2 derivatives checked
- mradermacher/chatty-djinn-14B-i1-GGUF: Quantization
- mradermacher/chatty-djinn-14B-GGUF: Quantization