GPU Task Manager (gputaskman) has been released in version v0.4.1.

The experimental, open-source terminal tool is an “htop for your GPU”: it shows per-process GPU memory and engine activity and tries to identify loaded LLMs with their memory breakdown. It runs on Linux (NVIDIA plus AMD and Intel), Windows, WSL and macOS.

Changes in version 0.4.1 (paraphrased):

• Detects custom model-path arguments and tracks where they came from (#49)
• Reports Intel GPU activity using Windows WDDM counters (#53)
• Shows NVIDIA GPUs in TCC mode on Windows (#37)
• Keeps Intel integrated WDDM memory out of VRAM reporting (#54)
• Reports recording I/O failures cleanly (#58)
• Identifies loaded Foundry Local models on Snapdragon (#62)
• Reports Qualcomm GPU memory as shared RAM (#63)
• Reads the Windows GPU driver version from PnP (#64)
• Fixes integrated GPU process memory reporting (#65)

Source: GitHub