{ "id": 583962, "title": "I need help with inference code", "url": "/competitions/cmi-detect-behavior-with-sensor-data/discussion/583962", "author_name": "", "post_date": "2025-06-10T16:43:20.384089200Z", "votes": null, "comment_count": 5, "views": 0, "content": "

My code fails in inference again and again. Here error: GatewayRuntimeError: (, '')

", "messages": [ { "id": "3221258", "postDate": "06/10/2025 16:43:20", "content": "

My code fails in inference again and again. Here error: GatewayRuntimeError: (, '')

", "rawMarkdown": "My code fails in inference again and again. Here error: \tGatewayRuntimeError: (, '')", "votes": null }, { "id": "3222176", "postDate": "06/12/2025 00:48:31", "content": "

You probably trained models in Inference Notebook, which will cause a timeout error. Upload your trained model to the inference notebook instead.

", "rawMarkdown": "You probably trained models in Inference Notebook, which will cause a timeout error. Upload your trained model to the inference notebook instead.", "votes": null }, { "id": "3222229", "postDate": "06/12/2025 03:37:03", "content": "

No, separately.

", "rawMarkdown": "No, separately.", "votes": null }, { "id": "3223613", "postDate": "06/13/2025 15:01:45", "content": "

can i see the code

", "rawMarkdown": "can i see the code", "votes": null }, { "id": "3224674", "postDate": "06/15/2025 09:55:18", "content": "

I encountered a similar issue, and it was resolved by explicitly returning the prediction result as a string. I haven't seen your code, so I'm not sure if this applies, but if it's worth trying, give it a shot.

\n
 (gesture)\n
", "rawMarkdown": "I encountered a similar issue, and it was resolved by explicitly returning the prediction result as a string. I haven't seen your code, so I'm not sure if this applies, but if it's worth trying, give it a shot.\n\n```python\nreturn str(gesture)\n```", "votes": null }, { "id": "3225822", "postDate": "06/16/2025 22:23:23", "content": "

I had the same issue as @Isao. The API error message(s) are not very helpful.

\n

My predict code was returning type np._str rather than str, which looks the same but causes problems with the API code. I have a cell in my notebook that makes sure the predict function is returning a str:

\n
 sequence_id  (test[]):\n\n    =predict(test.filter(test[] == sequence_id),\n              test_demographics)\n    (sequence_id,,p)\n\n    # Should be <class >:\n    (type(p))\n
", "rawMarkdown": "I had the same issue as @Isao. The API error message(s) are not very helpful.\n\nMy predict code was returning type `np._str` rather than `str`, which looks the same but causes problems with the API code. I have a cell in my notebook that makes sure the predict function is returning a `str`:\n\n```\nfor sequence_id in set(test['sequence_id']):\n \n p=predict(test.filter(test['sequence_id'] == sequence_id),\n test_demographics)\n print(sequence_id,':',p)\n \n # Should be :\n print(type(p))\n```", "votes": null } ], "comments": [ { "id": 3222176, "author_name": "shanzhong8", "author_url": "", "post_date": "06/12/2025 00:48:31", "content": "

You probably trained models in Inference Notebook, which will cause a timeout error. Upload your trained model to the inference notebook instead.

", "votes": null, "replies": [ { "id": 3222229, "author_name": "rustambazarbayev", "author_url": "", "post_date": "06/12/2025 03:37:03", "content": "

No, separately.

", "votes": null, "replies": [] } ] }, { "id": 3223613, "author_name": "ramanofthesky", "author_url": "", "post_date": "06/13/2025 15:01:45", "content": "

can i see the code

", "votes": null, "replies": [] }, { "id": 3224674, "author_name": "utm529fg", "author_url": "", "post_date": "06/15/2025 09:55:18", "content": "

I encountered a similar issue, and it was resolved by explicitly returning the prediction result as a string. I haven't seen your code, so I'm not sure if this applies, but if it's worth trying, give it a shot.

\n
 (gesture)\n
", "votes": null, "replies": [ { "id": 3225822, "author_name": "nnjjpp", "author_url": "", "post_date": "06/16/2025 22:23:23", "content": "

I had the same issue as @Isao. The API error message(s) are not very helpful.

\n

My predict code was returning type np._str rather than str, which looks the same but causes problems with the API code. I have a cell in my notebook that makes sure the predict function is returning a str:

\n
 sequence_id  (test[]):\n\n    =predict(test.filter(test[] == sequence_id),\n              test_demographics)\n    (sequence_id,,p)\n\n    # Should be <class >:\n    (type(p))\n
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "3221258": "My code fails in inference again and again. Here error: \tGatewayRuntimeError: (, '')", "3222176": "You probably trained models in Inference Notebook, which will cause a timeout error. Upload your trained model to the inference notebook instead.", "3222229": "No, separately.", "3223613": "can i see the code", "3224674": "I encountered a similar issue, and it was resolved by explicitly returning the prediction result as a string. I haven't seen your code, so I'm not sure if this applies, but if it's worth trying, give it a shot.\n\n```python\nreturn str(gesture)\n```", "3225822": "I had the same issue as @Isao. The API error message(s) are not very helpful.\n\nMy predict code was returning type `np._str` rather than `str`, which looks the same but causes problems with the API code. I have a cell in my notebook that makes sure the predict function is returning a `str`:\n\n```\nfor sequence_id in set(test['sequence_id']):\n \n p=predict(test.filter(test['sequence_id'] == sequence_id),\n test_demographics)\n print(sequence_id,':',p)\n \n # Should be :\n print(type(p))\n```" }, "source": "meta" }