Hi,
\n\nI have almost exhausted my TPU quota for this week and there is still lolot of time in reset. So, I was trying to use colab for training my models using TPU instance. It is not colab Pro nor do I have GCP account. I am using the free version at the moment. But, while trying to create TF dataset object I got an error. \nUnimplementedError: File system scheme '[local]' not implemented (file: '/content/01-5000.tfrec')\nIt is clear to me that we need a GCP bucket in order to feed the data. On Kaggle we can do so using KaggleDataset module. But on Colab we do not have such a module to use. I tried lot of googling and other things but none of them worked for me.
Any help would be appreciated. Thanks
", "messages": [ { "id": 920312, "postDate": "2020-07-08T13:47:37.617Z", "content": "Hi,
\n\nI have almost exhausted my TPU quota for this week and there is still lolot of time in reset. So, I was trying to use colab for training my models using TPU instance. It is not colab Pro nor do I have GCP account. I am using the free version at the moment. But, while trying to create TF dataset object I got an error. \nUnimplementedError: File system scheme '[local]' not implemented (file: '/content/01-5000.tfrec')\nIt is clear to me that we need a GCP bucket in order to feed the data. On Kaggle we can do so using KaggleDataset module. But on Colab we do not have such a module to use. I tried lot of googling and other things but none of them worked for me.
Any help would be appreciated. Thanks
", "rawMarkdown": "Hi,\n\nI have almost exhausted my TPU quota for this week and there is still lolot of time in reset. So, I was trying to use colab for training my models using TPU instance. It is not colab Pro nor do I have GCP account. I am using the free version at the moment. But, while trying to create TF dataset object I got an error. \n`UnimplementedError: File system scheme '[local]' not implemented (file: '/content/01-5000.tfrec')`\nIt is clear to me that we need a GCP bucket in order to feed the data. On Kaggle we can do so `using KaggleDataset` module. But on Colab we do not have such a module to use. I tried lot of googling and other things but none of them worked for me. \n\nAny help would be appreciated. Thanks", "votes": 3 }, { "id": 921070, "postDate": "2020-07-09T04:08:56.950Z", "content": "There is a way to do so. Here is the discussion link where you can find more details. https://www.kaggle.com/c/alaska2-image-steganalysis/discussion/165130#920639
\n\nThe idea is to give Kaggle Dataset GCS path in your colab notebook data and it will work. But, the links will expire after some time so be cautious about it.
", "rawMarkdown": "There is a way to do so. Here is the discussion link where you can find more details. [https://www.kaggle.com/c/alaska2-image-steganalysis/discussion/165130#920639](https://www.kaggle.com/c/alaska2-image-steganalysis/discussion/165130#920639)\n\nThe idea is to give Kaggle Dataset GCS path in your colab notebook data and it will work. But, the links will expire after some time so be cautious about it. ", "votes": 1 }, { "id": 924707, "postDate": "2020-07-11T15:32:32.360Z", "content": "but what is to stop someone uses multiple accounts to train on the competition, but only use one account to make submissions?
\n\nAlso, We know from previous experience that people used multiple accounts to create submissions, and then later joined multiple accounts into one team, and in the end they stayed on the LB
\n\nJust to make clear, not encouraging such behaviour, but that is what people are doing - just saying
", "rawMarkdown": "but what is to stop someone uses multiple accounts to train on the competition, but only use one account to make submissions?\n\nAlso, We know from previous experience that people used multiple accounts to create submissions, and then later joined multiple accounts into one team, and in the end they stayed on the LB\n\nJust to make clear, not encouraging such behaviour, but that is what people are doing - just saying", "replies": [ { "id": 924749, "postDate": "2020-07-11T15:59:37.800Z", "content": "That is true at some extent. But, we can't tackle all different scenarios. We can take part in competition and follow the rules. And can encourage others to do the same. That's best what a good competitor can do. Play and take part with good spirit.
", "rawMarkdown": "That is true at some extent. But, we can't tackle all different scenarios. We can take part in competition and follow the rules. And can encourage others to do the same. That's best what a good competitor can do. Play and take part with good spirit. " }, { "id": 924764, "postDate": "2020-07-11T16:11:05.793Z", "content": "anyone is entitled to use the platform in possible ways as long as they don't get DQed
\n\nalso pretty sure I am not a good competitor anymore :)
", "rawMarkdown": "anyone is entitled to use the platform in possible ways as long as they don't get DQed \n\nalso pretty sure I am not a good competitor anymore :) " }, { "id": 924804, "postDate": "2020-07-11T16:27:07.660Z", "content": "Why are not you good competitor anymore? 😂
", "rawMarkdown": "Why are not you good competitor anymore? 😂" } ] }, { "id": 922983, "postDate": "2020-07-10T13:12:47.247Z", "rawMarkdown": "", "votes": -3, "isDeleted": true, "replies": [ { "id": 923087, "postDate": "2020-07-10T14:21:29.120Z", "content": "You should know that multiple accounts are banned on Kaggle. That is against the rules.
", "rawMarkdown": "You should know that multiple accounts are banned on Kaggle. That is against the rules." }, { "id": 923169, "postDate": "2020-07-10T15:27:14.737Z", "content": "RIght, it is the first rule even.
", "rawMarkdown": "RIght, it is the first rule even." } ] } ], "comments": [ { "id": 921070, "author_name": "Urvish", "author_url": "", "post_date": "2020-07-09T04:08:56.950000", "content": "There is a way to do so. Here is the discussion link where you can find more details. https://www.kaggle.com/c/alaska2-image-steganalysis/discussion/165130#920639
\n\nThe idea is to give Kaggle Dataset GCS path in your colab notebook data and it will work. But, the links will expire after some time so be cautious about it.
", "votes": 1, "replies": [] }, { "id": 924707, "author_name": "Yifan Xie", "author_url": "", "post_date": "2020-07-11T15:32:32.360000", "content": "but what is to stop someone uses multiple accounts to train on the competition, but only use one account to make submissions?
\n\nAlso, We know from previous experience that people used multiple accounts to create submissions, and then later joined multiple accounts into one team, and in the end they stayed on the LB
\n\nJust to make clear, not encouraging such behaviour, but that is what people are doing - just saying
", "votes": 0, "replies": [ { "id": 924749, "author_name": "Urvish", "author_url": "", "post_date": "2020-07-11T15:59:37.800000", "content": "That is true at some extent. But, we can't tackle all different scenarios. We can take part in competition and follow the rules. And can encourage others to do the same. That's best what a good competitor can do. Play and take part with good spirit.
", "votes": 0, "replies": [] }, { "id": 924764, "author_name": "Yifan Xie", "author_url": "", "post_date": "2020-07-11T16:11:05.793000", "content": "anyone is entitled to use the platform in possible ways as long as they don't get DQed
\n\nalso pretty sure I am not a good competitor anymore :)
", "votes": 0, "replies": [] }, { "id": 924804, "author_name": "Urvish", "author_url": "", "post_date": "2020-07-11T16:27:07.660000", "content": "Why are not you good competitor anymore? 😂
", "votes": 0, "replies": [] } ] }, { "id": 922983, "author_name": "", "author_url": "", "post_date": "2020-07-10T13:12:47.247000", "content": "", "votes": -3, "replies": [ { "id": 923087, "author_name": "Urvish", "author_url": "", "post_date": "2020-07-10T14:21:29.120000", "content": "You should know that multiple accounts are banned on Kaggle. That is against the rules.
", "votes": 0, "replies": [] }, { "id": 923169, "author_name": "CPMP", "author_url": "", "post_date": "2020-07-10T15:27:14.737000", "content": "RIght, it is the first rule even.
", "votes": 0, "replies": [] } ] } ], "raw_markdown_by_id": { "920312": "Hi,\n\nI have almost exhausted my TPU quota for this week and there is still lolot of time in reset. So, I was trying to use colab for training my models using TPU instance. It is not colab Pro nor do I have GCP account. I am using the free version at the moment. But, while trying to create TF dataset object I got an error. \n`UnimplementedError: File system scheme '[local]' not implemented (file: '/content/01-5000.tfrec')`\nIt is clear to me that we need a GCP bucket in order to feed the data. On Kaggle we can do so `using KaggleDataset` module. But on Colab we do not have such a module to use. I tried lot of googling and other things but none of them worked for me. \n\nAny help would be appreciated. Thanks", "921070": "There is a way to do so. Here is the discussion link where you can find more details. [https://www.kaggle.com/c/alaska2-image-steganalysis/discussion/165130#920639](https://www.kaggle.com/c/alaska2-image-steganalysis/discussion/165130#920639)\n\nThe idea is to give Kaggle Dataset GCS path in your colab notebook data and it will work. But, the links will expire after some time so be cautious about it. ", "924707": "but what is to stop someone uses multiple accounts to train on the competition, but only use one account to make submissions?\n\nAlso, We know from previous experience that people used multiple accounts to create submissions, and then later joined multiple accounts into one team, and in the end they stayed on the LB\n\nJust to make clear, not encouraging such behaviour, but that is what people are doing - just saying", "922983": "" } }