While we're all waiting for the last few hours, having no time to do further experiments, I feel like this is an appropriate moment to re-share this blog post by @iglovikov, for those who haven't read it yet.
\nTLDR : Share your code and your solution, the more effort you put into it the more it will benefit you and others.
\nSome of the things explained here are actually quite advanced, I am usually too lazy to go further than step 4 (or 3), which I recommend doing as well as they are not really that time consuming if your code is not too dirty.
\nAlso, if you plan on doing a solution write-up, that's already most of the job done for your ReadMe… Which you can start working on now :)
", "messages": [ { "id": 1061228, "postDate": "2020-10-26T20:44:52.297Z", "content": "While we're all waiting for the last few hours, having no time to do further experiments, I feel like this is an appropriate moment to re-share this blog post by @iglovikov, for those who haven't read it yet.
\nTLDR : Share your code and your solution, the more effort you put into it the more it will benefit you and others.
\nSome of the things explained here are actually quite advanced, I am usually too lazy to go further than step 4 (or 3), which I recommend doing as well as they are not really that time consuming if your code is not too dirty.
\nAlso, if you plan on doing a solution write-up, that's already most of the job done for your ReadMe… Which you can start working on now :)
", "rawMarkdown": "While we're all waiting for the last few hours, having no time to do further experiments, I feel like this is an appropriate moment to re-share [this blog post](https://medium.com/kaggle-blog/i-trained-a-model-what-is-next-d1ba1c560e26) by @iglovikov, for those who haven't read it yet.\n\n**TLDR :** Share your code and your solution, the more effort you put into it the more it will benefit you and others.\n\nSome of the things explained here are actually quite advanced, I am usually too lazy to go further than step 4 (or 3), which I recommend doing as well as they are not really that time consuming if your code is not too dirty.\n\nAlso, if you plan on doing a solution write-up, that's already most of the job done for your ReadMe... Which you can start working on now :)", "votes": 9 }, { "id": 1061241, "postDate": "2020-10-26T21:06:15.047Z", "content": "For those you heavily use notebooks for coding, I would personally recommend https://nbdev.fast.ai/ which is a 3 in 1 kind of deal. Your notebook code being the main source automatically gets converted to a python library, documentation and testing suite by just making few minor syntax changes. Given the fact that our solution heavily relied on past competition winners I am pretty sure converting code into a library would benefit in the future competitions. Here is an example nbdev project I made to see what it looks like: https://github.com/KeremTurgutlu/self_supervised. Hope this helps easy modularized programming with some of the best practices like testing and documentation.
", "rawMarkdown": "For those you heavily use notebooks for coding, I would personally recommend https://nbdev.fast.ai/ which is a 3 in 1 kind of deal. Your notebook code being the main source automatically gets converted to a python library, documentation and testing suite by just making few minor syntax changes. Given the fact that our solution heavily relied on past competition winners I am pretty sure converting code into a library would benefit in the future competitions. Here is an example nbdev project I made to see what it looks like: https://github.com/KeremTurgutlu/self_supervised. Hope this helps easy modularized programming with some of the best practices like testing and documentation.", "votes": 3 } ], "comments": [ { "id": 1061241, "author_name": "Kerem Turgutlu", "author_url": "", "post_date": "2020-10-26T21:06:15.047000", "content": "For those you heavily use notebooks for coding, I would personally recommend https://nbdev.fast.ai/ which is a 3 in 1 kind of deal. Your notebook code being the main source automatically gets converted to a python library, documentation and testing suite by just making few minor syntax changes. Given the fact that our solution heavily relied on past competition winners I am pretty sure converting code into a library would benefit in the future competitions. Here is an example nbdev project I made to see what it looks like: https://github.com/KeremTurgutlu/self_supervised. Hope this helps easy modularized programming with some of the best practices like testing and documentation.
", "votes": 3, "replies": [] } ], "raw_markdown_by_id": { "1061228": "While we're all waiting for the last few hours, having no time to do further experiments, I feel like this is an appropriate moment to re-share [this blog post](https://medium.com/kaggle-blog/i-trained-a-model-what-is-next-d1ba1c560e26) by @iglovikov, for those who haven't read it yet.\n\n**TLDR :** Share your code and your solution, the more effort you put into it the more it will benefit you and others.\n\nSome of the things explained here are actually quite advanced, I am usually too lazy to go further than step 4 (or 3), which I recommend doing as well as they are not really that time consuming if your code is not too dirty.\n\nAlso, if you plan on doing a solution write-up, that's already most of the job done for your ReadMe... Which you can start working on now :)", "1061241": "For those you heavily use notebooks for coding, I would personally recommend https://nbdev.fast.ai/ which is a 3 in 1 kind of deal. Your notebook code being the main source automatically gets converted to a python library, documentation and testing suite by just making few minor syntax changes. Given the fact that our solution heavily relied on past competition winners I am pretty sure converting code into a library would benefit in the future competitions. Here is an example nbdev project I made to see what it looks like: https://github.com/KeremTurgutlu/self_supervised. Hope this helps easy modularized programming with some of the best practices like testing and documentation." } }