Parsing boto3 invoke_endpoint response from AWS SageMaker - parsing

Question

I have a Sagemaker endpoint that I can infer to from boto3 client and get response.
Per boto3 doc, the Body of the response result is a Byte object StreamingBody type. I convert it to a dictionary
response = client.invoke_endpoint(EndpointName=endpoint_name, Body=json.dumps(data))
response_body = response['Body']
dict_response = response_body.read().decode('utf-8')
print(dict_response)
The above code gives me a response like below (stripped down for this post)
I need to retrieve the array from the "floatVal" key. How do I do that?
{
"outputs": {
"score": {
"dtype": "DT_FLOAT",
"floatVal": [
0.00012408883776515722,
...........
-0.8316119909286499,
-0.24423488974571228
],
"tensorShape": {
"dim": [
{
"size": "1"
},
{
"size": "1024"
}
]
}
}
},
"modelSpec": {
"version": "1",
"name": "generic_model",
"signatureName": "serving_default"
}
}

Answer 1

Actually the dict_response is not really a dictionary here, rather a string type. So I had to convert the dict_response to an actual dictionary and then I could retrieve the floatVal key.
Updated code
response = client.invoke_endpoint(EndpointName=endpoint_name, Body=json.dumps(data))
response_body = response['Body']
response_str = response_body.read().decode('utf-8')
response_dict = eval(response_str)
print(response_dict['outputs']['score']['floatVal'])