Salastil/odysseus
1
0
Fork 0
mirror of https://github.com/pewdiepie-archdaemon/odysseus.git synced 2026-06-16 17:55:26 -04:00
Code Issues Packages Projects Releases Wiki Activity

fix(hwfit): detect NVIDIA GPU on WSL and other minimal-PATH environments (#3306)

Browse Source 
The nvidia-smi absolute-path fallback in _detect_nvidia() was gated
on _remote_host, so it never ran for local detection. On systems
where nvidia-smi is not in the default PATH (e.g. WSL: /usr/lib/wsl/lib/),
this caused the Cookbook to report 'No GPU' even when nvidia-smi works
from an interactive shell.

Two issues fixed:
1. Removed the _remote_host gate so the absolute-path scan runs for
   local detection too.
2. For local execution, pass arguments as a list instead of a string
   so subprocess.run() resolves the absolute path correctly. Remote
   (SSH) execution keeps the string form, which the SSH command builder
   handles.

Co-authored-by: Bipin Mishra <bipin.mishra@atlascopco.com>
This commit is contained in:
Bipin Mishra
2026-06-07 21:23:49 +05:30
committed by GitHub
parent a6bc1addd2
commit b22c2b280c
1 changed files with 13 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files
services/hwfit/hardware.py
+13 -5
View File
@@ -76,9 +76,10 @@ def _detect_nvidia():
global _last_gpu_error global _last_gpu_error
_last_gpu_error = None _last_gpu_error = None
out = _run(["nvidia-smi", "--query-gpu=memory.total,name", "--format=csv,noheader,nounits"]) out = _run(["nvidia-smi", "--query-gpu=memory.total,name", "--format=csv,noheader,nounits"])
# Remote fallback: a non-interactive SSH shell often has a minimal PATH # Fallback: a non-interactive shell (or WSL) often has a minimal PATH
# that omits where nvidia-smi lives (/usr/bin, /usr/local/cuda/bin), so the # that omits where nvidia-smi lives (/usr/bin, /usr/local/cuda/bin,
# first call silently returns nothing → "No GPU" on hosts that DO have GPUs. # /usr/lib/wsl/lib), so the first call silently returns nothing →
# "No GPU" on machines that DO have GPUs.
# Retry through a login shell with the common CUDA bin dirs on PATH. # Retry through a login shell with the common CUDA bin dirs on PATH.
if not out and _remote_host: if not out and _remote_host:
out = _run( out = _run(
@@ -88,9 +89,16 @@ def _detect_nvidia():
# Last resort: call nvidia-smi by absolute path. Some hosts have a login # Last resort: call nvidia-smi by absolute path. Some hosts have a login
# shell that isn't bash (or a profile that errors), so the bash -lc retry # shell that isn't bash (or a profile that errors), so the bash -lc retry
# above still comes back empty even though the binary is right there. # above still comes back empty even though the binary is right there.
if not out and _remote_host: # Also handles WSL where nvidia-smi lives at /usr/lib/wsl/lib/ — a path
# that may not be in the server process's PATH.
if not out:
for _p in ("/usr/bin/nvidia-smi", "/usr/local/bin/nvidia-smi", "/usr/local/cuda/bin/nvidia-smi", "/usr/lib/wsl/lib/nvidia-smi"): for _p in ("/usr/bin/nvidia-smi", "/usr/local/bin/nvidia-smi", "/usr/local/cuda/bin/nvidia-smi", "/usr/lib/wsl/lib/nvidia-smi"):
out = _run(f"{_p} --query-gpu=memory.total,name --format=csv,noheader,nounits") # Use list form so subprocess.run (local) resolves the absolute path
# correctly instead of treating the whole string as an executable name.
if _remote_host:
out = _run(f"{_p} --query-gpu=memory.total,name --format=csv,noheader,nounits")
else:
out = _run([_p, "--query-gpu=memory.total,name", "--format=csv,noheader,nounits"])
if out: if out:
break break
if not out: if not out: