![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F15426557%2F4a40c1552acd5c0e6b202abcddf12517%2FScreenshot%202024-10-25%20232620.png?generation=1729888059124016&alt=media\")
I'm sorry but what comment exactly are you trying to refer to?\nI thought that the comment in the pic is the final update(for now at least )
The guidelines say:
\n\"Each batch of predictions (except the very first) must be returned within 1 minute of the batch features being provided.\"
\nDoes this 1-minute rule mirror real-world needs? For example, self-driving cars and other time-sensitive applications need rapid predictions—certainly not models that take 10 minutes. This limit might push us to optimize models for practical, quick responses in real-time scenarios which can make it difficult to do enough features engineering🤕
\nTBH I think this competition lack some clarity đź’”
", "messages": [ { "id": "3027934", "postDate": "10/25/2024 12:59:19", "content": "The guidelines say:
\n\"Each batch of predictions (except the very first) must be returned within 1 minute of the batch features being provided.\"
\nDoes this 1-minute rule mirror real-world needs? For example, self-driving cars and other time-sensitive applications need rapid predictions—certainly not models that take 10 minutes. This limit might push us to optimize models for practical, quick responses in real-time scenarios which can make it difficult to do enough features engineering🤕
\nTBH I think this competition lack some clarity đź’”
", "rawMarkdown": "The guidelines say:\n\n\"Each batch of predictions (except the very first) must be returned within 1 minute of the batch features being provided.\"\n\nDoes this 1-minute rule mirror real-world needs? For example, self-driving cars and other time-sensitive applications need rapid predictions—certainly not models that take 10 minutes. This limit might push us to optimize models for practical, quick responses in real-time scenarios which can make it difficult to do enough features engineering🤕\n\nTBH I think this competition lack some clarity 💔", "votes": null }, { "id": "3027942", "postDate": "10/25/2024 13:11:12", "content": "\n\nThis limit might push us to optimize models for practical, quick responses in real-time scenarios which can make it difficult to do enough features engineering
\n
Welcome to the real world.
", "rawMarkdown": ">This limit might push us to optimize models for practical, quick responses in real-time scenarios which can make it difficult to do enough features engineering\n\n\nWelcome to the real world.", "votes": null }, { "id": "3027964", "postDate": "10/25/2024 13:44:02", "content": "@nandodmelo @aymanallawi can you corroborate the place where 1-minute condition is highlighted? I did not see an instruction regarding this on the overview and data page.
\nDid you infer this from the evaluation script?
", "rawMarkdown": "nandodmelo @aymanallawi can you corroborate the place where 1-minute condition is highlighted? I did not see an instruction regarding this on the overview and data page.\n\nDid you infer this from the evaluation script?", "votes": null }, { "id": "3027966", "postDate": "10/25/2024 13:51:35", "content": "In the notebook: Jane Street RMF Demo Submission
\nIn the second code cell you can find this comment: \"Each batch of predictions (except the very first) must be returned within 1 minute of the batch features being provided.\"
Don’t panic. The 1min is likely a typo.
\n", "rawMarkdown": "Don’t panic. The 1min is likely a typo.\n\nhttps://www.kaggle.com/competitions/jane-street-real-time-market-data-forecasting/discussion/542022#3027305", "votes": null }, { "id": "3028305", "postDate": "10/25/2024 20:27:17", "content": "I'm sorry but what comment exactly are you trying to refer to?
\nI thought that the comment in the pic is the final update(for now at least )
I thought it was 10 minutes? If its 1 minute it might completely rule out online learning..
", "rawMarkdown": "I thought it was 10 minutes? If its 1 minute it might completely rule out online learning..", "votes": null }, { "id": "3028325", "postDate": "10/25/2024 21:31:42", "content": "In the demo submission, it was written 10min time limit, until someone in the above discussion mentioned that the evaluation script actually hard-coded 60 seconds for the time limit. The question there was whether the 60s was a typo. Sohier couldn't confirm yet so he simply changed the description in the Demo to make sure it aligns with the code. Ryan is the final guy who confirms it whether 1min or 10min.
\nSo dont panic yet. Wait for Ryan finally confirms.
In all probability, yes? There is a minimum of 849 time_id per day, which if you presume is around a minute per interval, and if I understand it correctly, each predict call yields only 1 time_id, then there are 14 hours in that.
![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F17516663%2F99d6db450dc42fade6c0dde72bdf4c2f%2F2024-11-10%20113633.png?generation=1731209856044361&alt=media\")
it's 60s, line 14 of kaggle_evaluation/jane_street_gateway.py: self.set_response_timeout_seconds(60)
\n\nThis limit might push us to optimize models for practical, quick responses in real-time scenarios which can make it difficult to do enough features engineering
\n
Welcome to the real world.
", "votes": null, "replies": [] }, { "id": 3027964, "author_name": "ravi20076", "author_url": "", "post_date": "10/25/2024 13:44:02", "content": "@nandodmelo @aymanallawi can you corroborate the place where 1-minute condition is highlighted? I did not see an instruction regarding this on the overview and data page.
\nDid you infer this from the evaluation script?
", "votes": null, "replies": [ { "id": 3027966, "author_name": "aymanallawi", "author_url": "", "post_date": "10/25/2024 13:51:35", "content": "In the notebook: Jane Street RMF Demo Submission
\nIn the second code cell you can find this comment: \"Each batch of predictions (except the very first) must be returned within 1 minute of the batch features being provided.\"
Don’t panic. The 1min is likely a typo.
\n", "votes": null, "replies": [ { "id": 3028305, "author_name": "aymanallawi", "author_url": "", "post_date": "10/25/2024 20:27:17", "content": "I'm sorry but what comment exactly are you trying to refer to?
\nI thought that the comment in the pic is the final update(for now at least )
In the demo submission, it was written 10min time limit, until someone in the above discussion mentioned that the evaluation script actually hard-coded 60 seconds for the time limit. The question there was whether the 60s was a typo. Sohier couldn't confirm yet so he simply changed the description in the Demo to make sure it aligns with the code. Ryan is the final guy who confirms it whether 1min or 10min.
\nSo dont panic yet. Wait for Ryan finally confirms.
I thought it was 10 minutes? If its 1 minute it might completely rule out online learning..
", "votes": null, "replies": [ { "id": 3092911, "author_name": "alexmason11", "author_url": "", "post_date": "01/10/2025 08:12:11", "content": "it's 60s, line 14 of kaggle_evaluation/jane_street_gateway.py: self.set_response_timeout_seconds(60)
In all probability, yes? There is a minimum of 849 time_id per day, which if you presume is around a minute per interval, and if I understand it correctly, each predict call yields only 1 time_id, then there are 14 hours in that.