I followed the Dataflow Compiler Tutorials and wanted to convert my self-trained object detection model to .hef. The previous steps were completed smoothly, and I validated both the Native and Quantized models. However, the conversion failed during the Compilation Tutorial, and the documentation does not provide detailed explanations, which has left me quite confused. How can I convert my own model, or where can I find relevant documentation for guidance?
If you have successfully quantized your model, the compilation should run through in most cases.
If you encounter an error, please share the error message you received.
You can find the documentation in the Developer Zone. Additionally, when you start the Hailo AI Software Suite Docker the first time, a directory containing the documentation in PDF format will be created at:
/local/shared_with_docker/doc
here is the error message, can you give me some suggestions.
[info] Loading network parameters
[info] Starting Hailo allocation and compilation flow
[info] Finding the best partition to contexts…
[…<==>…] Elapsed: 00:00:37
[error] Mapping Failed (allocation time: 46s)
Compiler could not find a valid partition to contexts. Most common error is: Validator failed on node: format_conversion5_sdc with Agent infeasible
with 2/2 failures.
[error] Failed to produce compiled graph
[error] BackendAllocatorException: Compilation failed: Compiler could not find a valid partition to contexts. Most common error is: Validator failed on node: format_conversion5_sdc with Agent infeasible
with 2/2 failures.
Did you solved it. I got the same problem
My model is YOLOX, previously I merged several outputs into a single output which caused a compilation failure. After modifying the outputs, the issue was resolved.