mirror of
https://github.com/snapotter-hq/SnapOtter.git
synced 2026-08-03 07:46:42 +02:00
ashim-hq
37277e5c09
- Add safe_onnx_session() to gpu.py with graceful CUDA EP → CPU fallback - Replace bare ort.InferenceSession() calls across colorize, restore, inpaint, remove_bg - Add libcublas-12-6 to production Dockerfile for ONNX Runtime CUDA EP - Add skipIfFeatureNotInstalled guards to remove-bg, blur-faces, smart-crop, ocr, noise-removal e2e specs - Add AI tool install prompt detection in tools-all.spec.ts - Add smart-crop to PYTHON_SIDECAR_TOOLS so frontend shows install prompt correctly - Create Dockerfile.test.dockerignore to include tests/ in test image builds - Add libheif-examples and exiftool to Dockerfile.test for HEIC and metadata tests - Regenerate visual regression baselines for Docker/Linux and skip on non-Docker platforms
75 lines
2.8 KiB
Python
75 lines
2.8 KiB
Python
"""Runtime GPU/CUDA detection utility."""
|
|
import functools
|
|
import os
|
|
import subprocess
|
|
import sys
|
|
|
|
|
|
@functools.lru_cache(maxsize=1)
|
|
def gpu_available():
|
|
"""Return True if a usable CUDA GPU is present at runtime."""
|
|
# Allow explicit disable via env var (set to "false" or "0")
|
|
override = os.environ.get("ASHIM_GPU")
|
|
if override is not None and override.lower() in ("0", "false", "no"):
|
|
return False
|
|
|
|
# Use torch.cuda as the source of truth when available. It actually
|
|
# probes the hardware. Fall back to onnxruntime provider detection
|
|
# when torch is not installed (e.g. CPU-only images without PyTorch).
|
|
try:
|
|
import torch
|
|
avail = torch.cuda.is_available()
|
|
if avail:
|
|
name = torch.cuda.get_device_name(0)
|
|
print(f"[gpu] CUDA available via torch: {name}", file=sys.stderr, flush=True)
|
|
else:
|
|
print("[gpu] torch loaded but CUDA not available", file=sys.stderr, flush=True)
|
|
return avail
|
|
except ImportError as e:
|
|
print(f"[gpu] torch not importable: {e}", file=sys.stderr, flush=True)
|
|
|
|
# Fallback: check if onnxruntime-gpu is installed and CUDA EP is available,
|
|
# then verify an actual NVIDIA GPU is present via nvidia-smi.
|
|
try:
|
|
import onnxruntime as _ort
|
|
providers = _ort.get_available_providers()
|
|
if "CUDAExecutionProvider" not in providers:
|
|
return False
|
|
# CUDA EP is compiled in — verify hardware is actually present.
|
|
# nvidia-smi is the most reliable cross-platform check.
|
|
result = subprocess.run(
|
|
["nvidia-smi", "--query-gpu=name", "--format=csv,noheader"],
|
|
capture_output=True, text=True, timeout=5,
|
|
)
|
|
if result.returncode == 0 and result.stdout.strip():
|
|
print(f"[gpu] CUDA available via ONNX Runtime + nvidia-smi: {result.stdout.strip()}",
|
|
file=sys.stderr, flush=True)
|
|
return True
|
|
return False
|
|
except (ImportError, FileNotFoundError, subprocess.TimeoutExpired):
|
|
return False
|
|
|
|
|
|
def onnx_providers():
|
|
"""Return ONNX Runtime execution providers in priority order."""
|
|
if gpu_available():
|
|
return ["CUDAExecutionProvider", "CPUExecutionProvider"]
|
|
return ["CPUExecutionProvider"]
|
|
|
|
|
|
def safe_onnx_session(model_path, providers=None):
|
|
"""Create an ONNX Runtime InferenceSession with graceful CUDA EP fallback."""
|
|
import onnxruntime as ort
|
|
|
|
if providers is None:
|
|
providers = onnx_providers()
|
|
|
|
try:
|
|
return ort.InferenceSession(model_path, providers=providers)
|
|
except Exception as e:
|
|
if "CUDAExecutionProvider" in providers:
|
|
print(f"[gpu] CUDA EP init failed ({e}), falling back to CPU",
|
|
file=sys.stderr, flush=True)
|
|
return ort.InferenceSession(model_path, providers=["CPUExecutionProvider"])
|
|
raise
|