JX Coder 7B Agent Behavior is a Parameter-Efficient Fine-Tuning adapter developed by JumpLander for controlled software-engineering agents. — Base model: Qwen/Qwen2.5-Coder-7B-Instruct — Fine-tuning method: 4-bit QLoRA / PEFT — Primary dataset: jumplander/JL-AgentBehavior-10K — Primary language: English — Artifact type: LoRA adapter — Primary purpose: coding-agent behavioral policy — Developer: JumpLander This repository contains the trained adapter weights, not a standalone copy of the full 7B base model. At inference time, the adapter is loaded on top of Qwen2.5-Coder-7B-Instruct: The small adapter file size is expected. The base model provides general language and coding capability, while the JX adapter modifies the model toward a more controlled agent policy. > A behavioral-policy warm-start for software-engineering agents. — a model trained from random initialization; — a fully autonomous coding agent; — a runtime-verified repository repair model; — a replacement for repository execution; — a standalone benchmark winner; — a fully bilingual English–Persian model. The adapter is developed and fine-tuned by JumpLander, while the underlying language-model architecture and base…
Модальности:
Генерация текста
Области применения:
Диалог / чат Вызов функций (Tool use) Генерация кода
Задача: Генерация текста
Автор: jumplander
Теги: peft, jumplander, jx, qwen2.5, qwen2.5-coder, coding-agent, agentic-ai, software-engineering
Лайков: 4 | Загрузок: 6
Описание основано на материалах HuggingFace. Перевод выполнен автоматически.
