MercanAI/Mercan-0.8B-DPO-Q4

Mercan 0.8B DPO — Q4_K_M

This repository contains the DPO-final Mercan 0.8B deployment artifact, converted from the final DPO checkpoint before the later GRPO/RLVR stage.

Main artifact

• Mercan-0.8B-DPO-Q4_K_M.mercan — Mercan v1 / GGUF v3, Q4_K_M
• Mercan-0.8B-DPO-Q4_K_M.mercan.sha256 — SHA-256
• Mercan-0.8B-DPO-Q4_K_M.mercan.json — conversion/provenance manifest

Source training chain: continuation SFT -> DPO
Source checkpoint: step_00007895.pt

Model contract

• tokenizer: NDSRF004
• tokenizer SHA-256: 72412d981dac65a29d1767bc98821fc2bcffc2de53c534e7c719598515bfb600
• vocab rows: 32,002
• chat control IDs: 32,000 / 32,001
• canonical roles: sistem, kullanici, asistan
• hidden size: 1536
• layers: 24
• attention heads: 12
• KV heads: 4
• FFN width: 5632
• context/sliding window: 32,768
• RoPE base: 1,000,000
• MorphFFN layers: 18

The structural control IDs are stored with the Mercan runtime aliases <|im_start|> / <|im_end|> while role headers remain Turkish.