{"repo":"reacher-z/gpu-monitor","free":true,"listed":false,"github":"https://github.com/reacher-z/gpu-monitor","clone":"git clone https://github.com/reacher-z/gpu-monitor.git","description":"Lightweight NVIDIA GPU monitor — alerts on Slack/Discord/Telegram/20 channels when training crashes or GPU overheats. Zero dependencies. Single file.","language":"Python","stars":36,"topics":["deep-learning","gpu","gpu-monitoring","machine-learning","monitoring","nvidia","prometheus","pytorch","slack","alerting"],"license":"MIT","category":"machine-learning","readme_excerpt":"gpu-monitor — GPU crash and OOM alerting, zero dependencies Your training crashed at 3AM. Six hours of wasted compute. You find out in the morning. gpu-monitor catches it the moment it happens and alerts you — before hours of compute are wasted. One Python file. Zero dependencies. 20 notification channels. Three lines. You're covered. ⭐ Star it if it saves your next run — helps other ML engineers find the tool. --- Table of Contents - What Happens When... - Quick Start - Example Output - Why gpu-monitor? - Features - Supported Notification Channels - Environment Variables - General - Per-channel variables - Prometheus Metrics - Alertmanager Webhook Receiver - Kubernetes - GitHub Pages Dashboard - Multi-Machine Setup - Setting Up Specific Channels - Telegram - WeCom, Feishu, DingTalk, Bark - ntfy - Apprise (80+ extra services) - OpenClaw - Who Uses gpu-monitor? - Privacy and Security - Troubleshooting - Citing gpu-monitor - Author --- What Happens When... Real scenarios gpu-monitor handles automatically: Your training job crashes at 3 AM: You wake up to this Slack message and restart within minutes, instead of discovering 8 lost GPU-hours in the morning. A GPU overheats during a long run: You get paged before hardware damage or thermal throttling silently ruins your results. Memory is quietly leaking across epochs: Caught before you OOM-crash at epoch 47 and lose 6 hours of checkpoints. One GPU goes idle while others are busy (hung worker): Without this alert, a single stuck D","default_branch":null,"files":null,"tree":[],"storefront":"/r/reacher-z","claimed":false,"request_supported":{"post":"https://gitbuyer.com/r/reacher-z/gpu-monitor/request-supported","requests":0},"note":"indexed from public GitHub; nothing is for sale on this page. Clone it from GitHub. Paid listings live at /search."}