90a6ee6fbf
Full-stack app with FastAPI backend (SSH/paramiko, pipeline streaming, GPU stats, xterm.js terminal, Ollama model manager) and React + Tailwind frontend (8 panels: Connection, Documents, Pipeline, QA Pairs, Training, Terminal, Models, Config). Docker Compose included. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
40 lines
1.4 KiB
Python
40 lines
1.4 KiB
Python
from ssh_client import ssh_manager
|
|
|
|
|
|
def get_gpu_stats() -> dict:
|
|
"""Query nvidia-smi on the remote host and return parsed GPU info."""
|
|
try:
|
|
if not ssh_manager.is_connected():
|
|
return {"gpus": [], "error": "Not connected"}
|
|
|
|
out, err, code = ssh_manager.execute(
|
|
"nvidia-smi --query-gpu=name,utilization.gpu,memory.used,memory.total,"
|
|
"temperature.gpu,power.draw --format=csv,noheader,nounits",
|
|
use_conda=False
|
|
)
|
|
|
|
if code != 0:
|
|
return {"gpus": [], "error": err.strip() or "nvidia-smi failed"}
|
|
|
|
gpus = []
|
|
for line in out.strip().split("\n"):
|
|
if not line.strip():
|
|
continue
|
|
parts = [p.strip() for p in line.split(",")]
|
|
if len(parts) >= 5:
|
|
try:
|
|
gpus.append({
|
|
"name": parts[0],
|
|
"utilization": int(parts[1]),
|
|
"memory_used": int(parts[2]),
|
|
"memory_total": int(parts[3]),
|
|
"temperature": int(parts[4]),
|
|
"power_draw": float(parts[5]) if len(parts) > 5 else None,
|
|
})
|
|
except (ValueError, IndexError):
|
|
pass
|
|
|
|
return {"gpus": gpus, "error": None}
|
|
except Exception as e:
|
|
return {"gpus": [], "error": str(e)}
|