ENTITY PROFILE / 实体档案
Qwen3.5-0.8B
Fine‑tuning Qwen3.5‑0.8B raised decision accuracy from 37 % to 65 % over 60 steps, requiring about 10 minutes on 4 GB VRAM. Qwen‑2.5‑7B training uses veRL as the trainer Qwen‑2.5‑7B training is orchestrated by Ray Qwen‑2.5‑7B training is orchestrated by KubeRay Qwen‑2.5‑7B uses vLLM for rollout generation· Graphiti 生成的实体摘要
RELATIONS
4关系CURRENT
4当前有效EVENTS
1相关事件REPORTS
1来源报道暂无可公开的邻近关系
按需展开实体的全部已公开关系，保留探索过的节点与连线。
关系变化
按生效时间- Qwen3.5-0.8BORCHESTRATED BYRay
Qwen‑2.5‑7B training is orchestrated by Ray
- Qwen3.5-0.8BORCHESTRATED BYKubeRay
Qwen‑2.5‑7B training is orchestrated by KubeRay
- Qwen3.5-0.8BUSES TRAINERverl
Qwen‑2.5‑7B training uses veRL as the trainer
- Qwen3.5-0.8BROLLOUT WITHvLLM
Qwen‑2.5‑7B uses vLLM for rollout generation
关系是 Graphiti 从报道中提取的事实陈述，不是原文引述；失效时间表示后来的报道显示这条关系已不成立。