fancyfeast/llama-joycaption-alpha-two-hf-llava
Full fine-tune vs google/siglip-so400m-patch14-384. Fine-tune: 93% of the weights changed, across 26 of 27 layers.
Tensors changed: Embeddings 3/3, Norms 104/112, MLP 104/112, Attention 208/220, other 0/1
6 derivatives checked
- John6666/llama-joycaption-beta-one-hf-llava-nf4: Quantization
- Jobaar/Llama-JoyCaption-Alpha-Two-GGUF: Same weights, converted (fewer layers)
- JKCHSTR/llama-joycaption-alpha-two-hf-llava-FP8-Dynamic: Quantization
- razvanab/llama-joycaption-alpha-two-hf-llava-Q4_K_M-GGUF: Quantization (fewer layers)
- John6666/llama-joycaption-alpha-two-hf-llava-nf4: Quantization
- razvanab/llama-joycaption-alpha-two-hf-llava-Q8_0-GGUF: Quantization (fewer layers)
7 other repos hold byte-for-byte identical weight files.