I am new to kaggle, but not ML. There might be another discussion board for this question. Lmk.
\n\nWhat is the right workflow for these competitions on kaggle? Do folks download the datasets and go for it on their own systems?
\n\nOr do people use notebooks on kaggle itself? If so, what kind of machine are these notebooks running on?
\n\nWhat wud u advise?
", "messages": [ { "id": "439206", "postDate": "12/14/2018 22:21:46", "content": "I am new to kaggle, but not ML. There might be another discussion board for this question. Lmk.
\n\nWhat is the right workflow for these competitions on kaggle? Do folks download the datasets and go for it on their own systems?
\n\nOr do people use notebooks on kaggle itself? If so, what kind of machine are these notebooks running on?
\n\nWhat wud u advise?
", "rawMarkdown": "I am new to kaggle, but not ML. There might be another discussion board for this question. Lmk.\n\nWhat is the right workflow for these competitions on kaggle? Do folks download the datasets and go for it on their own systems? \n\nOr do people use notebooks on kaggle itself? If so, what kind of machine are these notebooks running on? \n\nWhat wud u advise?", "votes": null }, { "id": "439323", "postDate": "12/15/2018 06:32:51", "content": "I am also new to the Kaggle kernels. I just noticed tonight, the notebook/kernel could be super slow (never finished) for a simple task, such as printing the first 10 words in the Glove embedding model. Not sure if that's due to my stupid mis-usage or not.
", "rawMarkdown": "I am also new to the Kaggle kernels. I just noticed tonight, the notebook/kernel could be super slow (never finished) for a simple task, such as printing the first 10 words in the Glove embedding model. Not sure if that's due to my stupid mis-usage or not.", "votes": null }, { "id": "439353", "postDate": "12/15/2018 08:21:02", "content": "No worries @takeseven the community is here to help. We were also beginners at some point.
\n\n\n\n\nWhat is the right workflow for these competitions on kaggle? Do folks download the datasets and go for it on their own systems?
\n
This is kernels competition. So, we use kernels to run our scripts and make submission. Although few uses their own local build to tune the model or run hyperparameter search to get best values. And, then we use those values in kernels, run them and submit the score.
\n\nUsually for this competition a GPU and 16-32 Gigs of RAM is enough IMO. And, others just rely on kernels and run several experiments on kernels and use the best of them. I belong to that group of users.
", "rawMarkdown": "No worries @takeseven the community is here to help. We were also beginners at some point.\n\n> What is the right workflow for these competitions on kaggle? Do folks download the datasets and go for it on their own systems?\n\nThis is kernels competition. So, we use kernels to run our scripts and make submission. Although few uses their own local build to tune the model or run hyperparameter search to get best values. And, then we use those values in kernels, run them and submit the score. \n\nUsually for this competition a GPU and 16-32 Gigs of RAM is enough IMO. And, others just rely on kernels and run several experiments on kernels and use the best of them. I belong to that group of users.", "votes": null }, { "id": "444383", "postDate": "12/23/2018 23:10:09", "content": "Thanks. So, just working on Kaggle website works for many users. That's what I will do too. Is there a way to have multiple code files on Kaggle for the sample project?
", "rawMarkdown": "Thanks. So, just working on Kaggle website works for many users. That's what I will do too. Is there a way to have multiple code files on Kaggle for the sample project?", "votes": null }, { "id": "444482", "postDate": "12/24/2018 05:08:51", "content": "You can make serveral kernel notebooks/scripts for same competition.
", "rawMarkdown": "You can make serveral kernel notebooks/scripts for same competition.", "votes": null } ], "comments": [ { "id": 439323, "author_name": "feynmann", "author_url": "", "post_date": "12/15/2018 06:32:51", "content": "I am also new to the Kaggle kernels. I just noticed tonight, the notebook/kernel could be super slow (never finished) for a simple task, such as printing the first 10 words in the Glove embedding model. Not sure if that's due to my stupid mis-usage or not.
", "votes": null, "replies": [] }, { "id": 439353, "author_name": "shaz13", "author_url": "", "post_date": "12/15/2018 08:21:02", "content": "No worries @takeseven the community is here to help. We were also beginners at some point.
\n\n\n\n\nWhat is the right workflow for these competitions on kaggle? Do folks download the datasets and go for it on their own systems?
\n
This is kernels competition. So, we use kernels to run our scripts and make submission. Although few uses their own local build to tune the model or run hyperparameter search to get best values. And, then we use those values in kernels, run them and submit the score.
\n\nUsually for this competition a GPU and 16-32 Gigs of RAM is enough IMO. And, others just rely on kernels and run several experiments on kernels and use the best of them. I belong to that group of users.
", "votes": null, "replies": [ { "id": 444383, "author_name": "takeseven", "author_url": "", "post_date": "12/23/2018 23:10:09", "content": "Thanks. So, just working on Kaggle website works for many users. That's what I will do too. Is there a way to have multiple code files on Kaggle for the sample project?
", "votes": null, "replies": [] }, { "id": 444482, "author_name": "shaz13", "author_url": "", "post_date": "12/24/2018 05:08:51", "content": "You can make serveral kernel notebooks/scripts for same competition.
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "439206": "I am new to kaggle, but not ML. There might be another discussion board for this question. Lmk.\n\nWhat is the right workflow for these competitions on kaggle? Do folks download the datasets and go for it on their own systems? \n\nOr do people use notebooks on kaggle itself? If so, what kind of machine are these notebooks running on? \n\nWhat wud u advise?", "439323": "I am also new to the Kaggle kernels. I just noticed tonight, the notebook/kernel could be super slow (never finished) for a simple task, such as printing the first 10 words in the Glove embedding model. Not sure if that's due to my stupid mis-usage or not.", "439353": "No worries @takeseven the community is here to help. We were also beginners at some point.\n\n> What is the right workflow for these competitions on kaggle? Do folks download the datasets and go for it on their own systems?\n\nThis is kernels competition. So, we use kernels to run our scripts and make submission. Although few uses their own local build to tune the model or run hyperparameter search to get best values. And, then we use those values in kernels, run them and submit the score. \n\nUsually for this competition a GPU and 16-32 Gigs of RAM is enough IMO. And, others just rely on kernels and run several experiments on kernels and use the best of them. I belong to that group of users.", "444383": "Thanks. So, just working on Kaggle website works for many users. That's what I will do too. Is there a way to have multiple code files on Kaggle for the sample project?", "444482": "You can make serveral kernel notebooks/scripts for same competition." }, "source": "meta" }