Search results

上下文会不断累加，最终超过模型的context length然后报错

Status: Open.

#23 In jd-opensource/JoyAI-VL-Interaction;

· zhangxuejian-a opened on Jul 2, 2026

about SFT

Status: Open.

#22 In jd-opensource/JoyAI-VL-Interaction;

· xuchanghua-88 opened on Jul 1, 2026

asr只是识别浏览器麦克风音频，在rtsp模式下缺失音频识别功能

Status: Open.

#21 In jd-opensource/JoyAI-VL-Interaction;

· romantic2016cool opened on Jul 1, 2026

4大模型对机器配置要求有点高了，而且只推理一个摄像头，实际落地场景，一个摄像头很难捕捉全有用信息，后续这方面有什么优化计划吗

Status: Open.

#20 In jd-opensource/JoyAI-VL-Interaction;

· romantic2016cool opened on Jul 1, 2026

可以建一个即时公共平台，方便大家讨论吗，比如微信群？

Status: Open.

#19 In jd-opensource/JoyAI-VL-Interaction;

· romantic2016cool opened on Jun 30, 2026

rtsp首次推理比较慢，建议添加VLLM预热、解码器预热功能

Status: Open.

#18 In jd-opensource/JoyAI-VL-Interaction;

· romantic2016cool opened on Jun 30, 2026

后训练脚本有计划开源吗？想针对自己的场景做一些后训练

Status: Open.

#17 In jd-opensource/JoyAI-VL-Interaction;

· SpareNoEfforttt opened on Jun 30, 2026

使用摄像头的RTSP拉流传输视频，限制了16K的上下文（只占用A6000单卡30多G显存），持续监控几分钟后显存占用超出大小

Status: Open.

#16 In jd-opensource/JoyAI-VL-Interaction;

· 18398814190 opened on Jun 30, 2026

5090 32跑不动啊

Status: Open.

#15 In jd-opensource/JoyAI-VL-Interaction;

· xqy99482-source opened on Jun 29, 2026

基模为什么选择Qwen3+Qwen3-VL ViT的结构

Status: Open.

#14 In jd-opensource/JoyAI-VL-Interaction;

· psq512 opened on Jun 29, 2026

数据集构建的工具有计划开源吗，可以制作自己场景的数据进行finetune？

Status: Open.

#13 In jd-opensource/JoyAI-VL-Interaction;

· mmxuan18 opened on Jun 27, 2026

JoyAI-VL-Interaction-Preview produces incoherent outputs in both vLLM and Transformers

Status: Open.

#12 In jd-opensource/JoyAI-VL-Interaction;

· Coco422 opened on Jun 24, 2026