何文松
|
7c129674c0
fix: 修复设备选择逻辑,确保 GPU 环境下使用 gpu:0
|
há 1 dia atrás |
何文松
|
386d5b0359
feat: 添加 GPU 显存检测和 MinerU 服务自动控制
|
há 1 dia atrás |
何文松
|
8c60c709d6
refactor: 使用独立脚本调用 PaddleOCR,避免显存共享问题
|
há 1 dia atrás |
何文松
|
5c5a032fbd
fix: 修改 call_paddleocr 函数使用 Python API
|
há 1 dia atrás |
何文松
|
4c4a7c4acb
refactor: 移除 PaddleOCR 命令行调用,只使用 Python API
|
há 1 dia atrás |
何文松
|
deaa0bfefe
feat: 添加 PaddleOCR Python API 支持(暂时禁用,显存不足)
|
há 1 dia atrás |
何文松
|
de7b25c053
feat: 使用 PaddleOCR Python API 替代命令行方式,支持图表识别和纯文本识别
|
há 1 dia atrás |
何文松
|
0282fe550c
Revert "fix: 未配置 VL 后端时使用传统 ocr 命令,避免加载 VL 模型导致 OOM"
|
há 1 dia atrás |
何文松
|
415a260763
fix: 未配置 VL 后端时使用传统 ocr 命令,避免加载 VL 模型导致 OOM
|
há 1 dia atrás |
何文松
|
600bdb85e2
fix: PaddleOCR命令自动检测venv路径 + 添加PDF OCR测试
|
há 1 semana atrás |
何文松
|
c8d3f04f05
feat: 识别异常时用Paddle解析全文档(full_document+extract_all_pages_from_pdf)
|
há 3 semanas atrás |
何文松
|
e69dff9ab7
feat: 检测MinerU识别异常(同字重复)时用Paddle doc_parser结果替换markdown再解析
|
há 3 semanas atrás |
何文松
|
81e98c0a90
fix: 备用解析时内容为图片但扩展名为.pdf则复制为正确扩展名再调doc_parser,避免PDFium Data format error
|
há 3 semanas atrás |
何文松
|
2dd570737c
chore: 移除 PaddleOCR 子进程 LD_PRELOAD/static TLS 逻辑
|
há 3 semanas atrás |
何文松
|
974d87f967
chore: 日志中区分图表识别与文本识别([PaddleOCR 图表识别] / [PaddleOCR 文本识别])
|
há 3 semanas atrás |
何文松
|
f6c245facc
refactor: 将 call_paddleocr_ocr 改为使用不识别图表的 doc_parser 替代 ocr 子命令
|
há 3 semanas atrás |
何文松
|
ed94d6102e
fix: 修复 PaddleOCR ocr 命令不支持 VL 参数的问题
|
há 3 semanas atrás |
何文松
|
160834c486
feat: 适配全项目 PaddleOCR 命令行以支持 VL 识别后端配置
|
há 3 semanas atrás |
何文松
|
1cd66b8826
feat: 为 PaddleOCR doc_parser 添加 VL 识别后端配置支持
|
há 3 semanas atrás |
何文松
|
692a0a4103
refactor: 优化配置文件并恢复部分底层环境变量读取逻辑
|
há 3 semanas atrás |
何文松
|
080d9e4463
feat: 实现基于 YAML/JSON 的统一配置文件系统
|
há 3 semanas atrás |
何文松
|
3e478f6b42
清理项目:删除多余的测试文件和重复文档
|
há 3 semanas atrás |
何文松
|
0fe830c65a
fix(paddleocr): 子进程注入 LD_PRELOAD 与 PADDLE_PDX 避免 static TLS 与模型源检查
|
há 4 semanas atrás |
何文松
|
554cf82e2b
pdf_converter_v2: GPU/NPU 采集适配、Paddle/MinerU 多卡单任务用满
|
há 4 semanas atrás |
何文松
|
14d0f42f6d
pdf_converter_v2: 移除停止 mineru-api.service 的逻辑及开关
|
há 4 semanas atrás |
何文松
|
3f0d1df186
pdf_converter_v2: 添加 MINERU_RELEASE_BEFORE_PADDLE_OCR 开关,可选不释放 MinerU
|
há 4 semanas atrás |
何文松
|
2aba4a8a3b
fix(ocr): resolve paddleocr executable path for systemd/venv (PADDLEOCR_CMD, same-dir as sys.executable)
|
há 1 mês atrás |
何文松
|
206bdccbb4
pdf_converter_v2: 同步设备环境识别(nvi/npu)、mineru-api.service、config/utils;mineru: models_download_utils local配置None检查
|
há 1 mês atrás |
何文松
|
d04debc556
pdf_converter_v2: 同步 GitLab/Clerk2.5 修改
|
há 1 mês atrás |
何文松
|
bf107371e7
完善适配NPU
|
há 1 mês atrás |