| NedoLM 0.8B SFT checkpoint: step_00006478.pt | |
| GGUF SHA256: 154542c2ca5c6ab310f80fb3582d1055f0753cbb8787a6aad418dfcc6b88b5e1 | |
| Tokenizer: NDSRF004 | |
| Tokenizer SHA256: 72412d981dac65a29d1767bc98821fc2bcffc2de53c534e7c719598515bfb600 | |
| GGUF public tokenizer metadata: tokenizer.ggml.model=rwkv (for Hugging Face GGUF parser compatibility) | |
| Exact runtime tokenizer selection: general.architecture=nedolm -> NDSRF004 Rust ABI | |
| llama.cpp base commit: e71b80510c848c00175924ecf3c40333ccae8eb5 | |
| NDSRF004 ID validation: job 1551320, NDSRF004_ID_MATCH_OK | |
| Inference smoke: job 1551321, COMPLETED 0:0 | |
| Architecture requires the NedoLM-patched llama.cpp backend (MorphFFN + SWA + NeoX RoPE + exact NDSRF004). | |