初始化项目，由ModelHub XC社区提供模型

Model: pathcosmos/frankenstallm
Source: Original Platform

gguf/Modelfile.3b-Q4_K_M

@@ -0,0 +1,26 @@
+# FRANKENSTALLM 3B Korean ORPO v1 — Ollama Modelfile
+# Usage:
+#   ollama create frankenstallm-3b -f Modelfile.3b
+#   ollama run frankenstallm-3b
+#
+# Sampling config verified via ORPO eval grid + Ollama 실측:
+#   PyTorch: temp=0.7, rep_pen=1.2 → rep=0%, EOS=100%
+#   Ollama Q4_K_M: 3-gram rep=1.8% (자연 어절 반복), EOS=100%
+#   greedy 대비 30.89%→1.8%로 퇴행적 반복 완전 해소
+
+FROM ./outputs/gguf/frankenstallm-3b-Q4_K_M.gguf
+
+# --- Sampling Config (ORPO eval grid + Ollama 검증) ---
+PARAMETER temperature 0.7
+PARAMETER repeat_penalty 1.2
+PARAMETER top_p 0.9
+PARAMETER top_k 50
+PARAMETER num_predict 512
+PARAMETER num_ctx 4096
+PARAMETER stop "</s>"
+
+# --- System Prompt ---
+SYSTEM """당신은 FRANKENSTALLM, 한국어에 특화된 3B 파라미터 언어 모델입니다. 정확하고 자연스러운 한국어로 답변해주세요."""
+
+# --- License ---
+LICENSE """Apache-2.0"""