<?xml version="1.0" ?><entry xml:lang="en-us" xmlns="http://www.w3.org/2005/Atom" xmlns:planet="http://planet.intertwingly.net/"><id>https://blog.nelhage.com/post/pickles-and-ml/</id><link href="https://blog.nelhage.com/post/pickles-and-ml/" rel="alternate" type="text/html"/><title>What's with ML software and pickles?</title><summary>I have spent many years as an software engineer who was a total outsider to machine-learning, but with some curiosity and occasional peripheral interactions with it. During this time, a recurring theme for me was horror (and, to be honest, disdain) every time I encountered the widespread usage of Python pickle in the Python ML ecosystem.
In addition to their major security issues1, the use of pickle for serialization tends to be very brittle, leading to all kinds of nightmares as you evolve your code and upgrade libraries and Python versions.</summary><updated planet:format="November 08, 2023 05:00 AM">2023-11-08T05:00:00Z</updated><source><id>https://blog.nelhage.com/post/</id><author><name>Nelson Elhage</name></author><link href="https://blog.nelhage.com/post/" rel="alternate" type="text/html"/><link href="https://blog.nelhage.com/post/atom.xml" rel="self" type="application/rss+xml"/><subtitle>Recent content in Posts on Made of Bugs</subtitle><title>Posts on Made of Bugs</title><updated planet:format="September 15, 2026 05:37 AM">2026-09-15T05:37:44Z</updated><planet:format>rss20</planet:format><planet:bozo>false</planet:bozo><planet:items_per_page>60</planet:items_per_page><planet:name>Nelson Elhage</planet:name><planet:css-id>nelson-elhage</planet:css-id><planet:days_per_page>0</planet:days_per_page><planet:http_etag>&quot;5b2ae9eec2e293572f6ae5a000bd0161-ssl-df&quot;</planet:http_etag><planet:http_status>200</planet:http_status><planet:encoding>utf-8</planet:encoding></source></entry>