mirror of
https://github.com/forkless/XPUSYS-Monitor-NG.git
synced 2026-08-16 00:46:37 +02:00
27 lines
1.1 KiB
Plaintext
27 lines
1.1 KiB
Plaintext
# CPU / RAM monitoring — required by all providers
|
|
psutil
|
|
|
|
# NVIDIA GPU monitoring — required by NvidiaProvider only
|
|
# Safe to install on Intel/AMD machines; unused if no NVIDIA driver is present
|
|
pynvml
|
|
|
|
# AMD GPU monitoring (Linux) — commented out for Windows-native ROCm
|
|
# rocm_smi_lib is Linux-only; Windows ROCm uses torch.cuda APIs instead.
|
|
# Install only if you are on Linux with ROCm and want driver-level GPU load:
|
|
# pip install rocm_smi_lib
|
|
# rocm_smi_lib
|
|
|
|
# ---------------------------------------------------------------------------
|
|
# The following are provided by ComfyUI itself — DO NOT add them here.
|
|
# Listed for reference only so manual environment setups know the minimums.
|
|
# ---------------------------------------------------------------------------
|
|
#
|
|
# torch >= 2.5
|
|
# torch.xpu (Intel Arc) was officially introduced in PyTorch 2.5.
|
|
# torch.cuda (NVIDIA/AMD ROCm) has been stable since PyTorch 1.x.
|
|
# Install the ROCm-enabled build from: https://pytorch.org/get-started/locally/
|
|
#
|
|
# aiohttp >= 3.8
|
|
# Used for HTTP routes and WebSocket broadcast.
|
|
# ComfyUI already depends on aiohttp; no separate installation needed.
|