ERROR: from element .../GstVvas_XInfer:vvas_xinfer0: infer config file parse error
Failed to set pipeline to PAUSED.
This error is preceded by a specific validation message indicating the root cause:
a) Invalid value for inference-backend-runtime:
ERROR vvas_xinfer gstvvas_xinfer_helper.cpp:read_infer_config:<vvas_xinfer0>
Invalid value [invalid_runtime] provided for inference-backend-runtime
should be one of [onnxrt, vart]
Possible fix - inference-backend-runtime only accepts onnxrt or vart:
"infer-config": {
"inference-backend-runtime": "onnxrt"
}
b) Unknown model-format:
ERROR vvas_xinfer gstvvas_xinfer_helper.cpp:read_infer_config:<vvas_xinfer0>
Unknown model-format <value> : supported values are [RGB, BGR]
c) ONNXRT configuration not found or invalid:
ERROR vvas_xinfer gstvvas_xinfer_helper.cpp:read_infer_config:<vvas_xinfer0>
ONNXRT configuration not found or invalid
Possible fix - when inference-backend-runtime is onnxrt, the onnxrt-config block is mandatory. The following shows a minimal example - adjust paths and parameters to match your model:
"infer-config": {
"inference-backend-runtime": "onnxrt",
"onnxrt-config": {
"onnx-model-path": "/etc/vai/models/ResNet50_INT8/ResNet50_INT8.onnx",
"input-tensor-layout": "NCHW",
"onnxrt-ep": "vitisai",
"vitisai-ep-config": {
"config-file-path": "/etc/vai/models/ResNet50_INT8/vitisai_config.json",
"cache-dir": "/etc/vai/models",
"cache-key": "ResNet50_INT8"
}
}
}