yzhao062/pyod

Cannot save AutoEncoder

Open

#256 建立於 2020年12月6日

在 GitHub 查看
 (6 留言) (1 反應) (1 負責人)Python (7,762 star) (1,308 fork)batch import
buggood first issuehelp wanted

描述

The official instructions say to use joblib for pickling PyOD models.

This fails for AutoEncoders, or any other TensorFlow-backed model as far as I can tell. The error is:

>>> dump(model, 'model.joblib')
...
TypeError: can't pickle _thread.RLock objects

Note that it's not sufficient to save the underlying Keras Sequential model, since I need the methods & variables of BaseDetector (like .decision_scores_ or .decision_function().

貢獻者指南

技術棧
python
領域
machine learning
議題類型
bug
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。
half day
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。
stale
清晰度議題是否清楚說明預期改動、驗收標準和下一步。
clear
前置要求
PythonTensorFlowjoblib
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。
30
研究方向
Investigate the AutoEncoder class in pyod/models/auto encoder.py. The error occurs because TensorFlow models contain non picklable objects. Consider implementing a custom save method that saves the Keras model weights and the detector's decision scores separately, then restores them on load. Check the existing model persistence documentation and any previous discussions in issue #256 for suggested approaches.