Error Handling¶

Common errors, solutions, and debugging techniques.

🎯 Overview¶

Learn how to handle and debug common issues.

🐛 Common Errors¶

Model Not Found¶

Error: FileNotFoundError: Model file not found

Solution:

# Check file exists
ls -lh model.onnx

# Use absolute path
gpux build /full/path/to/project

Input Validation Failed¶

Error: Input mismatch. Missing: {'input'}

Solution:

# Check input names
gpux inspect model-name

# Match input names in JSON
{
  "correct_input_name": [data]
}

Shape Mismatch¶

Error: Shape mismatch: expected [1,10], got [1,5]

Solution:

# Check expected shape
gpux inspect model

# Provide correct shape
data = np.zeros((1, 10))

Provider Not Available¶

Error: RuntimeError: No execution providers available

Solution:

# Install GPU runtime
pip install onnxruntime-gpu

# Or fallback to CPU
gpux build . --provider cpu

Out of Memory¶

Error: RuntimeError: CUDA out of memory

Solution:

# Reduce memory limit
runtime:
  gpu:
    memory: 1GB
  batch_size: 1

🔍 Debugging¶

Enable Verbose Logging¶

gpux run model --verbose

Check Model Details¶

gpux inspect model --json > info.json
cat info.json

Test with CPU¶

# Rule out GPU issues
gpux build . --provider cpu
gpux run model --input data.json

Validate Configuration¶

python -c "from gpux.config.parser import GPUXConfigParser; GPUXConfigParser().parse_file('gpux.yml')"

🛡️ Exception Handling¶

Python¶

from gpux import GPUXRuntime

try:
    runtime = GPUXRuntime("model.onnx")
    result = runtime.infer(data)
except FileNotFoundError:
    print("Model not found")
except ValueError:
    print("Invalid input data")
except RuntimeError as e:
    print(f"Runtime error: {e}")
finally:
    runtime.cleanup()

💡 Key Takeaways¶

Success

✅ Common errors and solutions ✅ Debugging techniques ✅ Exception handling ✅ Validation methods

Previous: Python API