Cannot convert JSON float numbers

[thomasSve]

Bug Report

System information

• OS Platform and Distribution: macOS 10
• TensorFlow Serving installed from (source or binary): Deployed from docker image tensorflow/serving
• TensorFlow Serving version: 1.11

Describe the problem

I believe there is an issue with how the json numbers are checked

https://github.com/tensorflow/serving/blob/master/tensorflow_serving/util/json_tensor.cc#L210

Example:
curl -X POST http://localhost:8501/v1/models/estimator:predict -H 'cache-control: no-cache' -d '{ "signature_name": "predict", "instances": [{"speed": 0.96 }] }'

returns:
{ "error": "Failed to process element: 0 key: speed of \'instances\' list. Error: Invalid argument: Cannot convert JSON value: 0.96 to float without loss of precision." }

---

Another example when testing the example in https://www.tensorflow.org/serving/api_rest#example

curl -d '{"instances": [1.0,2.0,5.0]}' -X POST http://localhost:8501/v1/models/half_plus_three:predict

returns fine:
{ "predictions": [3.5, 4.0, 5.5 ]

while changing one of the numbers, for example:
curl -d '{"instances": [1.24,5.0]}' -X POST http://localhost:8501/v1/models/half_plus_three:predict

returns:
{ "error": "Failed to process element: 0 of \'instances\' list. Error: Invalid argument: Cannot convert JSON value: 1.24 to float without loss of precision." }

[WarrenZevon]

I am seeing the same behavior on Windows 10, running Tensorflow Serving through a docker image. HTTP requests return successfully for floating point numbers with less than 2 numbers after the decimal point, but fail with more precision.

The error message returned:
{ "error": "Failed to process element: 0 of 'instances' list. Error: Invalid argument: Cannot convert JSON value: 234.04 to float without loss of precision." }