Troubleshooting / Bug Reporting

If you experience any bugs, we're happy to help as fast as possible! Please provide these details when reporting an issue:

1. Example code that reproduces the issue
2. What you expected to happen
3. What happened instead
4. Logs: You can increase log verbosity using the set_log_level function at the beginning of your script. The default level is "INFO".

Python

import denkflow
denkflow.set_log_level("DEBUG") # Or "TRACE" for maximum detail. Careful: TRACE output might include sensitive information.
# ... the rest of your script

C

#include "denkflow.h"
const char* log_level = "INFO";
DenkflowResult r = set_log_level(log_level);
// ... the rest of your script

5. Enable onnxruntime logs by setting the environment variable DENKFLOW_ENABLE_ORT_LOGS=true before running your application. These logs might contain valuable information.
6. System Information (Operating System, CPU Model, Memory, GPU Model if applicable).

We are actively developing the DENKflow API and welcome feature requests! Please feel free to report missing features or suggest improvements through the same channels you use for bug reporting.

Common Issues

• The first evaluation after a program's start will take a longer time than subsequent evaluations. This is especially noticeable when using CUDA or TensorRT. To avoid this phenomenon from interfering with benchmarks or cycle times in production, it is recommended to implement a test run before the main work cycle.