Docker image: nvcr.io/nvidia/deepstream-l4t:5.1-21.02-samples
input or output layers are empty, check trtis model config settings
ERROR: failed to initialize backend while setup model:ssd_inception_v2_coco_2018_01_28 layer info, nvinfer error:NVDSINFER_TRTIS_ERROR
config.pbtxt file:
name: “ssd_inception_v2_coco_2018_01_28”
platform: “tensorflow_graphdef”
default_model_filename: “model.graphdef”
max_batch_size: 128
input [
{
name: “image_tensor”
data_type: TYPE_UINT8
format: FORMAT_NHWC
dims: [ 300, 300, 3 ]
}
]
output [
{
name: “detection_boxes”
data_type: TYPE_FP32
dims: [ 100, 4]
reshape { shape: [100,4] }
},
{
name: “detection_classes”
data_type: TYPE_FP32
dims: [ 100 ]
},
{
name: “detection_scores”
data_type: TYPE_FP32
dims: [ 100 ]
},
{
name: “num_detections”
data_type: TYPE_FP32
dims: [ 1 ]
reshape { shape: }
}
]
The docker image starts with triton 2.5, it works fine with all models.
But after I download tritonserver2.10.0-jetpack4.5.tgz and extract it to reinstall triton servers, when I run deepstream-app, all deepstream file config returned same error “input or output layers are empty”.
Any idea to fix it? Please help. Thank you so much