Repository Issue Activity (beta)

#345 Verify evals on Papers with Code

#344 Improvement vectors for DeepSeek-OCR

#342 基于中间层隐状态读取的内生安全通路——替代外挂护栏与参数对齐的第三种架构

#26 prompt修改不起作用

#341 [教程TUTORIAL] DeepSeek-OCR Windows WSL本地安装部署避坑指南 Local Installation and Deployment Guide for Windows WSL

#340 Invitation: WDCC Shanghai subconference on the forgetting-mechanism direction

#190 Does it support multiple images per prompt?

#18 安装环境卡在 Building wheel for flash-attn (setup.py) 4个小时以上了~

#339 Hallucination/generation loop on Blackwell GPUs (sm_120)

#9 benchmark against mineru v2.5?

#240 [Solved] How I Got DeepSeek-OCR vLLM Working on an RTX 5090

#132 【想法探讨】讨论是否可行--使用“文本内容（现代汉语）→文言文”转换极致压缩上下文输入

#299 DeepSeek OCR Triton Error [CUDA] an illegal memory access on vLLM 0.11.2

#305 表格识别的问题

#327 能给出单个字符的矩形框吗（需要一个矩形框里最多一个字符，不能一个矩形框里多个字符）

#301 Query About the Evaluation

#329 📋 Documentation Enhancement Suggestion

#328 Is it only using 25% of the GPU for anyone else?

#325 是否有线上体验测试版 基于deepsseek-ocr模型的网页工具

#65 [Warning]: You are using a model of type deepseek_vl_v2 to instantiate a model of type DeepseekOCR. This is not supported for all configurations of models and can yield errors.

#318 How does Deepseek ocr compare to the new LightOnOCR-2-1B

#306 关于使用transformer 进行调用推理时，没能将结果保存到指定的目录下的问题。

#130 下载“vllm-0.8.5+cu118-cp38-abi3-manylinux1_x86_64.whl”时提示不存在

#156 ERROR: vllm-0.8.5+cu118-cp38-abi3-manylinux1_x86_64.whl is not a supported wheel on this platform.

#322 Vulnerability Discolure