Dear Organizer, would the \"Throughput phase\" be using the same data and be the same task (i.e also reconstruct the tracks, but focus on speed).
\n\nSome of my current python code are really slow and i am wondering if i should write some code in c++. If the code can be reused in \"Throughput phase\" , then it will be worth the effort to start some coding now.
\n\nThanks!
", "messages": [ { "id": "350580", "postDate": "06/30/2018 02:22:35", "content": "Dear Organizer, would the \"Throughput phase\" be using the same data and be the same task (i.e also reconstruct the tracks, but focus on speed).
\n\nSome of my current python code are really slow and i am wondering if i should write some code in c++. If the code can be reused in \"Throughput phase\" , then it will be worth the effort to start some coding now.
\n\nThanks!
", "rawMarkdown": "Dear Organizer, would the \"Throughput phase\" be using the same data and be the same task (i.e also reconstruct the tracks, but focus on speed).\n\nSome of my current python code are really slow and i am wondering if i should write some code in c++. If the code can be reused in \"Throughput phase\" , then it will be worth the effort to start some coding now.\n\nThanks!", "votes": null }, { "id": "350646", "postDate": "06/30/2018 06:36:08", "content": "Thé throughput phase will be the same task on the same dataset(the test dataset will be fresh events). The score will combine the accuracy score, the one used here, and the speed. So:\nInvestment in speeding up your code will not be lost. \nAlgorithms which are « naturally » fast but not yielding the highest accuracy now will have a second chance
", "rawMarkdown": "Thé throughput phase will be the same task on the same dataset(the test dataset will be fresh events). The score will combine the accuracy score, the one used here, and the speed. So:\nInvestment in speeding up your code will not be lost. \nAlgorithms which are « naturally » fast but not yielding the highest accuracy now will have a second chance", "votes": null }, { "id": "350647", "postDate": "06/30/2018 06:37:54", "content": "thanks for the reply!
", "rawMarkdown": "thanks for the reply!", "votes": null }, { "id": "350921", "postDate": "06/30/2018 18:51:00", "content": "This can help speed Python code: https://www.ibm.com/developerworks/community/blogs/jfp/entry/Python_Meets_Julia_Micro_Performance?lang=en
", "rawMarkdown": "This can help speed Python code: https://www.ibm.com/developerworks/community/blogs/jfp/entry/Python_Meets_Julia_Micro_Performance?lang=en", "votes": null }, { "id": "350991", "postDate": "07/01/2018 01:03:20", "content": "Thanks, May I ask how could you \"measure\" the speed? By using kernel here, or by any standard? Different computer has different specs, so I really wonder how the organiser evaluate this metric?
", "rawMarkdown": "Thanks, May I ask how could you \"measure\" the speed? By using kernel here, or by any standard? Different computer has different specs, so I really wonder how the organiser evaluate this metric?", "votes": null }, { "id": "351036", "postDate": "07/01/2018 06:05:34", "content": "It will not take place on kaggle but on Codalab. We have borrowed identical computers from the lhc computing grid. You will submit your code, we will run it automatically on a virtual machine single core 2GB memory. We measure the cpu time of events evaluation ( not counting overall initialisation or loading data in memory). The docker file which will include your code has all the usual libraries but you can load additional ones. \nThis is just a teaser, we ll release full information in a few days.
", "rawMarkdown": "It will not take place on kaggle but on Codalab. We have borrowed identical computers from the lhc computing grid. You will submit your code, we will run it automatically on a virtual machine single core 2GB memory. We measure the cpu time of events evaluation ( not counting overall initialisation or loading data in memory). The docker file which will include your code has all the usual libraries but you can load additional ones. \nThis is just a teaser, we ll release full information in a few days.", "votes": null }, { "id": "353453", "postDate": "07/06/2018 19:28:30", "content": "Haven't ran into this problem unfortunately... Maybe you can run the Docker container with GPUs or larger instances to accelerate iterations while you are still fine-tuning the model. Ive run Docker containers in the cloud using a tool I built called RunBox (https://www.tenzar.com/docs/concepts/overview). If you go this route, I can help you further.
", "rawMarkdown": "Haven't ran into this problem unfortunately... Maybe you can run the Docker container with GPUs or larger instances to accelerate iterations while you are still fine-tuning the model. Ive run Docker containers in the cloud using a tool I built called RunBox (https://www.tenzar.com/docs/concepts/overview). If you go this route, I can help you further.", "votes": null } ], "comments": [ { "id": 350646, "author_name": "droussea", "author_url": "", "post_date": "06/30/2018 06:36:08", "content": "Thé throughput phase will be the same task on the same dataset(the test dataset will be fresh events). The score will combine the accuracy score, the one used here, and the speed. So:\nInvestment in speeding up your code will not be lost. \nAlgorithms which are « naturally » fast but not yielding the highest accuracy now will have a second chance
", "votes": null, "replies": [ { "id": 350991, "author_name": "khahuras", "author_url": "", "post_date": "07/01/2018 01:03:20", "content": "Thanks, May I ask how could you \"measure\" the speed? By using kernel here, or by any standard? Different computer has different specs, so I really wonder how the organiser evaluate this metric?
", "votes": null, "replies": [] }, { "id": 351036, "author_name": "droussea", "author_url": "", "post_date": "07/01/2018 06:05:34", "content": "It will not take place on kaggle but on Codalab. We have borrowed identical computers from the lhc computing grid. You will submit your code, we will run it automatically on a virtual machine single core 2GB memory. We measure the cpu time of events evaluation ( not counting overall initialisation or loading data in memory). The docker file which will include your code has all the usual libraries but you can load additional ones. \nThis is just a teaser, we ll release full information in a few days.
", "votes": null, "replies": [] } ] }, { "id": 350647, "author_name": "hengck23", "author_url": "", "post_date": "06/30/2018 06:37:54", "content": "thanks for the reply!
", "votes": null, "replies": [ { "id": 353453, "author_name": "kevinyb", "author_url": "", "post_date": "07/06/2018 19:28:30", "content": "Haven't ran into this problem unfortunately... Maybe you can run the Docker container with GPUs or larger instances to accelerate iterations while you are still fine-tuning the model. Ive run Docker containers in the cloud using a tool I built called RunBox (https://www.tenzar.com/docs/concepts/overview). If you go this route, I can help you further.
", "votes": null, "replies": [] } ] }, { "id": 350921, "author_name": "cpmpml", "author_url": "", "post_date": "06/30/2018 18:51:00", "content": "This can help speed Python code: https://www.ibm.com/developerworks/community/blogs/jfp/entry/Python_Meets_Julia_Micro_Performance?lang=en
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "350580": "Dear Organizer, would the \"Throughput phase\" be using the same data and be the same task (i.e also reconstruct the tracks, but focus on speed).\n\nSome of my current python code are really slow and i am wondering if i should write some code in c++. If the code can be reused in \"Throughput phase\" , then it will be worth the effort to start some coding now.\n\nThanks!", "350646": "Thé throughput phase will be the same task on the same dataset(the test dataset will be fresh events). The score will combine the accuracy score, the one used here, and the speed. So:\nInvestment in speeding up your code will not be lost. \nAlgorithms which are « naturally » fast but not yielding the highest accuracy now will have a second chance", "350647": "thanks for the reply!", "350921": "This can help speed Python code: https://www.ibm.com/developerworks/community/blogs/jfp/entry/Python_Meets_Julia_Micro_Performance?lang=en", "350991": "Thanks, May I ask how could you \"measure\" the speed? By using kernel here, or by any standard? Different computer has different specs, so I really wonder how the organiser evaluate this metric?", "351036": "It will not take place on kaggle but on Codalab. We have borrowed identical computers from the lhc computing grid. You will submit your code, we will run it automatically on a virtual machine single core 2GB memory. We measure the cpu time of events evaluation ( not counting overall initialisation or loading data in memory). The docker file which will include your code has all the usual libraries but you can load additional ones. \nThis is just a teaser, we ll release full information in a few days.", "353453": "Haven't ran into this problem unfortunately... Maybe you can run the Docker container with GPUs or larger instances to accelerate iterations while you are still fine-tuning the model. Ive run Docker containers in the cloud using a tool I built called RunBox (https://www.tenzar.com/docs/concepts/overview). If you go this route, I can help you further." }, "source": "meta" }