Loading mode.json leads to Error: Corrupted Configuration

TensorFlow
,
1

Hi, I’m seeing an error when trying to load model.json in javascript. This error persists in the live server view as well as in the real version.

“Modified by moderator”
Unhandled Promise Rejection: Error: Corrupted configuration, expected array for nodeData: [object Object]


I'm expecting the model to load, but it won't load because the node weights aren't being correctly read. 

I've been able to run the web app using someone else's `model.json` and corresponding `.bin`  files, but my model throws this error every time. 

I've had quite a bit of trouble converting the .h5 in Jupyter to .json and .bin. There were several compatibiility issues that I somehow got around by installing older packages. In the end, it was converted while printing   "unable to determine model, this file likely only contains node weights.' but not in an "Error:" format.

I would like a solution or a workaround. I just want to use the model to predict handwritten digits. Thanks .
1 Like
3

Hi @KrC, This might be due to the model was not properly saved into the json file. Could you please try to save the model again and load it. If the error still persists could you please provide the standalone code to reproduce the issue. Thank You.