Hi, everyone, I'm a Kaggle novice, and it's the first time to join the compete, so I also have some questions about the rules.
\nFirstly, there is a saying that \"No internet access enabled on submission\", so it's OK to \"!pip install https://github.com/CellProfiling/HPA-Cell-Segmentation/archive/master.zip\" for cellsegmentator?
\nBesides, about TPU, \"TPUs will not be available for making submissions to this competition. You are still welcome to use them for training models.\" So, if I can train the model with TPU and generate the submission.csv in another Notebook?
\nThanks for your help! π
You will need to segment during prediction - in general for Kaggle it is OK in any competition to pip install libraries when internet access is not allowed. Several public data sets have been created by folks working on this competition that contain the additional libraries needed for this competition - you will need to use one of those data sets or create your own dataset that contains the library script (so as written your pip install code is not correct as your pointing to a github that will not be available with internet turned off).
\nYou may train your model any where with any device. The final prediction and submission needs to be performed in a kaggle notebook as that is the only way to \"see\" the private test data. So training a model in notebook using TPU's is perfect - the generated model is added to the prediction notebook in a data set. Kaggle makes it easy (but a tiny bit tricky) to attach a trained model from a different notebook.
", "rawMarkdown": "[GrayLee111](https://www.kaggle.com/graylee111)\n\nYou will need to segment during prediction - in general for Kaggle it is OK in any competition to pip install libraries when internet access is not allowed. Several public data sets have been created by folks working on this competition that contain the additional libraries needed for this competition - you will need to use one of those data sets or create your own dataset that contains the library script (so as written your pip install code is not correct as your pointing to a github that will not be available with internet turned off).\n\nYou may train your model any where with any device. The final prediction and submission needs to be performed in a kaggle notebook as that is the only way to \"see\" the private test data. So training a model in notebook using TPU's is perfect - the generated model is added to the prediction notebook in a data set. Kaggle makes it easy (but a tiny bit tricky) to attach a trained model from a different notebook.\n\n", "votes": 1, "replies": [ { "id": 1251644, "postDate": "2021-03-25T01:25:50.930Z", "content": "Itβs really hopeful for a novice! Thank you very much!π
", "rawMarkdown": "Itβs really hopeful for a novice! Thank you very much!π" } ] }, { "id": 1249741, "postDate": "2021-03-23T13:45:16.107Z", "content": "Hi, everyone, I'm a Kaggle novice, and it's the first time to join the compete, so I also have some questions about the rules.
\nFirstly, there is a saying that \"No internet access enabled on submission\", so it's OK to \"!pip install https://github.com/CellProfiling/HPA-Cell-Segmentation/archive/master.zip\" for cellsegmentator?
\nBesides, about TPU, \"TPUs will not be available for making submissions to this competition. You are still welcome to use them for training models.\" So, if I can train the model with TPU and generate the submission.csv in another Notebook?
\nThanks for your help! π
You will need to segment during prediction - in general for Kaggle it is OK in any competition to pip install libraries when internet access is not allowed. Several public data sets have been created by folks working on this competition that contain the additional libraries needed for this competition - you will need to use one of those data sets or create your own dataset that contains the library script (so as written your pip install code is not correct as your pointing to a github that will not be available with internet turned off).
\nYou may train your model any where with any device. The final prediction and submission needs to be performed in a kaggle notebook as that is the only way to \"see\" the private test data. So training a model in notebook using TPU's is perfect - the generated model is added to the prediction notebook in a data set. Kaggle makes it easy (but a tiny bit tricky) to attach a trained model from a different notebook.
", "votes": 1, "replies": [ { "id": 1251644, "author_name": "GrayLee111", "author_url": "", "post_date": "2021-03-25T01:25:50.930000", "content": "Itβs really hopeful for a novice! Thank you very much!π
", "votes": 0, "replies": [] } ] } ], "raw_markdown_by_id": { "1249873": "[GrayLee111](https://www.kaggle.com/graylee111)\n\nYou will need to segment during prediction - in general for Kaggle it is OK in any competition to pip install libraries when internet access is not allowed. Several public data sets have been created by folks working on this competition that contain the additional libraries needed for this competition - you will need to use one of those data sets or create your own dataset that contains the library script (so as written your pip install code is not correct as your pointing to a github that will not be available with internet turned off).\n\nYou may train your model any where with any device. The final prediction and submission needs to be performed in a kaggle notebook as that is the only way to \"see\" the private test data. So training a model in notebook using TPU's is perfect - the generated model is added to the prediction notebook in a data set. Kaggle makes it easy (but a tiny bit tricky) to attach a trained model from a different notebook.\n\n", "1249741": "Hi, everyone, I'm a Kaggle novice, and it's the first time to join the compete, so I also have some questions about the rules.\nFirstly, there is a saying that \"*No internet access enabled on submission*\", so it's OK to \"!pip install https://github.com/CellProfiling/HPA-Cell-Segmentation/archive/master.zip\" for cellsegmentator?\nBesides, about TPU, \"*TPUs will not be available for making submissions to this competition. You are still welcome to use them for training models.*\" So, if I can train the model with TPU and generate the submission.csv in another Notebook?\nThanks for your help! π\n" } }