“Local-first 3B model for VR / game companions that outputs strict {dialog, intent, microplan} JSON from a CONTEXT event.” English-only finetune of Qwen2.5-3B-Instruct for intent + microplan–driven NPC dialog. The model reads a structured CONTEXT JSON (environment, relationship, mood, signals) and produces: intent (one of 19 whitelisted labels) microplan (low-level action primitives) dialog as strict JSON** > v2 = refinement of v1: cleaned & rebalanced dataset, tighter JSON guardrails, and improved persona adherence. v2 is more stable (almost no JSON leaks), better label alignment, and more consistent diegetic tone. LoRA adapters (PEFT, SFT) → checkpoints/adapterfinal Merged FP16 → ./ GGUF quants (llama.cpp / llama-cpp-python) → gguf/sft-q6k.gguf, gguf/sft-q4km.gguf GGUF Style Fine-tune (Example) → gguf/rinstyle.gguf` (See fine-tuning section) These are the “sweet spot” sampling settings used in the Unity client (Ollama/llama.cpp-style). They balance creativity with JSON stability for Rin: Max Resample: 2 Resample Temp Step: 0.1 Memory: last 10 dialog turns + 6` recent actions You can safely lower temperature to ~0.7 if you want less playful dialog, or disable Mirostat…
Модальности:
Генерация текста
Области применения:
Диалог / чат
Задача: Генерация текста
Автор: AndriLawrence
Теги: gguf, qwen2, qwen, qwen2.5, 3b, lora, peft, sft
Лайков: 4 | Загрузок: 634
Описание основано на материалах HuggingFace. Перевод выполнен автоматически.
