How can i get the right label from a yolov5 based TensorRT model in Triton Server?

AI & Data Science Deep Learning (Training & Inference) Frameworks
,
1

As the title i mentioned,i send a gPRC request to my Triton Server successfully,but i got a super weried response,it`s not what i expect string with label,instead is a weried encoded Python string like this:

[[b'680.980530:226010' b'657.416748:224994' b'654.782410:225830'
  b'652.399475:225174']]

To figure out why it return like that,here is my Python client side code below:

outputs = [ 
    grpcclient.InferRequestedOutput('output0',class_count=4), 
    ]
    inputs = [grpcclient.InferInput('images', img.shape, 'FP32')] 
    inputs[0].set_data_from_numpy(img) 
    result = triton_client.infer( 
    'action_trt', 
    inputs, 
    request_id=str('1'),   # randomly input
    model_version='', 
    outputs=outputs) 
    #got server response
    output_array = result.as_numpy('output0')[0] # change it into numpy array
    print(output_array)

And here is my config.pbtxt like this：

name: "action_trt"
platform: "tensorrt_plan"
max_batch_size: 1
input [
  {
    name: "images",
    data_type: TYPE_FP32,
    dims: [ 3,640,640 ],
    
  }
]
output[
  {
    name: "output0",
    data_type: TYPE_FP32
    dims: [ 25200,9 ]
    label_filename: "labels.txt"
  }
]
version_policy: { latest { num_versions: 1 } }
instance_group: [
  {
    count: 1
    kind: KIND_GPU
    gpus: [ 0 ]
  }
]

So how can i fix that,since i have a school projecct close to ddl,it`s really need “Cardiotonic” for me and let me make a hot rush.
Huge thanks and regards
Elin Liu