meditsolutions/Llama-3.2-SUN-2.5B-chat
Partial fine-tune (fewer layers) vs meta-llama/Llama-3.2-1B-Instruct. meta-llama/Llama-3.2-1B-Instruct matches part of meditsolutions/Llama-3.2-SUN-2.5B-chat (146 of its 291 tensors; the rest of meditsolutions/Llama-3.2-SUN-2.5B-chat has no counterpart). On the shared part: Partial change: 43% of the weights changed (attention, embeddings, mlp, norms), in 6 of 16 layers.
Tensors changed: Embeddings 1/1, Norms 8/33, MLP 13/48, Attention 21/64
4 derivatives checked
- mradermacher/Llama-3.2-SUN-2.5B-chat-i1-GGUF: Full fine-tune
- mradermacher/Llama-3.2-SUN-2.5B-chat-GGUF: Partial fine-tune
- mkurman/Llama-3.2-MedIT-SUN-2.5B-BT-GRPO: Full fine-tune
- tensorblock/Llama-3.2-SUN-2.5B-chat-GGUF: Quantization