pytorch/serve

Handle invalid input intelligently ！

Open

#2,459 建立於 2023年7月12日

在 GitHub 查看
 (7 留言) (0 反應) (0 負責人)Java (3,844 star) (790 fork)batch import
debugginggood first issuetriaged

描述

🚀 The feature

Given a custom service that expects JSON as input, If the input is invalid JSON such as { "image": "I missed a quote} it will return { "code": 500, "type": "InternalServerException", "message": "Worker died." } and reload model. Is something can do to verify the accuracy of input to avoid such direct returns and model reload when it cannot be modified input?

Motivation, pitch

Is something can do to verify the accuracy of input to avoid such direct returns and model reload when it cannot be modified input?

Alternatives

No response

Additional context

No response

貢獻者指南

技術棧
pythonjavarest api
領域
backendapimachine learning
議題類型
feature
難度面向新貢獻者的預計實作難度，1 表示很小改動，5 表示專家級工作。
3
預計時間有經驗貢獻者完成調查、實作、測試並準備 pull request 的粗略時間範圍。
1-3 hours
活動狀態議題目前的可參與程度：新鮮、活躍、陳舊、阻塞或等待維護者輸入。
fresh
清晰度議題是否清楚說明預期改動、驗收標準和下一步。
mostly clear
前置要求
Familiarity with PyTorch ServeBasic understanding of JSON parsing
新手友善度1-100 的估計分數，表示該議題對首次貢獻者的友善程度。
65
研究方向
Investigate the input validation pipeline in PyTorch Serve. Look at the request handler code in `frontend/server/src/main/java/org/pytorch/serve/http/InferenceRequestHandler.java` and the worker process in `backend/python/worker.py`. The issue suggests adding pre validation of JSON input before passing to the model worker. Consider implementing a JSON schema validation step in the frontend to catch invalid JSON early and return a 400 error instead of a 500. Check if there are existing input validation utilities in the codebase that can be leveraged or extended.