{
"id": 556153,
"title": "Please help with NN loading and prediction!!",
"url": "/competitions/jane-street-real-time-market-data-forecasting/discussion/556153",
"author_name": "",
"post_date": "2025-01-11T12:24:02.916193500Z",
"votes": null,
"comment_count": 2,
"views": 0,
"content": "I am training my own NN on my own device, the NN is built with Keras, the versions are as follow ![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F15001593%2F0e88d09acdbe47b890a4d6fd83c2c3e7%2FScreenshot%202025-01-11%20at%208.18.26pm.png?generation=1736598207203219&alt=media\")
It's different from the keras and tf version in Kaggle. However I keep getting this error when I load the model in Kaggle and try to make the predictions: ![\"error\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F15001593%2F2aad301b50862b44af7cf10ad65fc706%2FScreenshot%202025-01-11%20at%208.16.48pm.png?generation=1736598181887175&alt=media\")
, this error always happen after 30-40 minutes of prediction using the test api from one of the amazing notebook I found in the discussion. I Googled but still face this issue, I am using customised data generator to train the model and make the prediction, I wonder is it the keras version issue? or some other problems. It's really annoying as I've got so many NN to test the performance :( Thanks in advance for any suggestions. !!!!
",
"messages": [
{
"id": "3093864",
"postDate": "01/11/2025 12:24:02",
"content": "I am training my own NN on my own device, the NN is built with Keras, the versions are as follow It's different from the keras and tf version in Kaggle. However I keep getting this error when I load the model in Kaggle and try to make the predictions: , this error always happen after 30-40 minutes of prediction using the test api from one of the amazing notebook I found in the discussion. I Googled but still face this issue, I am using customised data generator to train the model and make the prediction, I wonder is it the keras version issue? or some other problems. It's really annoying as I've got so many NN to test the performance :( Thanks in advance for any suggestions. !!!!
",
"rawMarkdown": "I am training my own NN on my own device, the NN is built with Keras, the versions are as follow  It's different from the keras and tf version in Kaggle. However I keep getting this error when I load the model in Kaggle and try to make the predictions: , this error always happen after 30-40 minutes of prediction using the test api from one of the amazing notebook I found in the discussion. I Googled but still face this issue, I am using customised data generator to train the model and make the prediction, I wonder is it the keras version issue? or some other problems. It's really annoying as I've got so many NN to test the performance :( Thanks in advance for any suggestions. !!!!",
"votes": null
},
{
"id": "3094244",
"postDate": "01/11/2025 19:36:06",
"content": "How are you saving and loading the model?
",
"rawMarkdown": "How are you saving and loading the model?",
"votes": null
},
{
"id": "3094378",
"postDate": "01/12/2025 02:22:29",
"content": "@redfoongus Hi, thanks for the reply, below is how I save and load the model: ![\"save\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F15001593%2F1c562784b1d1568a8146d305fadadba6%2FScreenshot%202025-01-12%20at%2010.19.30am.png?generation=1736648439664663&alt=media\")
![\"load\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F15001593%2F40d003243f27041a0d704fed92255c48%2FScreenshot%202025-01-12%20at%2010.18.21am.png?generation=1736648455607215&alt=media\")
I've also reinstalled the tensorflow and keras in my own device yesterday to make sure the versions align with the kaggle environment, but the issue remains and I still get error when making the submission after 1 hour or so.
",
"rawMarkdown": "redfoongus Hi, thanks for the reply, below is how I save and load the model:   I've also reinstalled the tensorflow and keras in my own device yesterday to make sure the versions align with the kaggle environment, but the issue remains and I still get error when making the submission after 1 hour or so.",
"votes": null
}
],
"comments": [
{
"id": 3094244,
"author_name": "redfoongus",
"author_url": "",
"post_date": "01/11/2025 19:36:06",
"content": "How are you saving and loading the model?
",
"votes": null,
"replies": [
{
"id": 3094378,
"author_name": "denniswang07",
"author_url": "",
"post_date": "01/12/2025 02:22:29",
"content": "@redfoongus Hi, thanks for the reply, below is how I save and load the model: I've also reinstalled the tensorflow and keras in my own device yesterday to make sure the versions align with the kaggle environment, but the issue remains and I still get error when making the submission after 1 hour or so.
",
"votes": null,
"replies": []
}
]
}
],
"raw_markdown_by_id": {
"3093864": "I am training my own NN on my own device, the NN is built with Keras, the versions are as follow  It's different from the keras and tf version in Kaggle. However I keep getting this error when I load the model in Kaggle and try to make the predictions: , this error always happen after 30-40 minutes of prediction using the test api from one of the amazing notebook I found in the discussion. I Googled but still face this issue, I am using customised data generator to train the model and make the prediction, I wonder is it the keras version issue? or some other problems. It's really annoying as I've got so many NN to test the performance :( Thanks in advance for any suggestions. !!!!",
"3094244": "How are you saving and loading the model?",
"3094378": "redfoongus Hi, thanks for the reply, below is how I save and load the model:   I've also reinstalled the tensorflow and keras in my own device yesterday to make sure the versions align with the kaggle environment, but the issue remains and I still get error when making the submission after 1 hour or so."
},
"source": "meta"
}