Just out of curiosity... Was wondering how strong is the correlation between computing power and LB score.\nMine is e5-1650, 16gb ram, 128gb ssd, 1tb sata, titan X (maxwell)
", "messages": [ { "id": "443092", "postDate": "12/21/2018 02:14:39", "content": "Just out of curiosity... Was wondering how strong is the correlation between computing power and LB score.\nMine is e5-1650, 16gb ram, 128gb ssd, 1tb sata, titan X (maxwell)
", "rawMarkdown": "Just out of curiosity... Was wondering how strong is the correlation between computing power and LB score.\nMine is e5-1650, 16gb ram, 128gb ssd, 1tb sata, titan X (maxwell)", "votes": null }, { "id": "443687", "postDate": "12/22/2018 06:32:31", "content": "ryzen 5 1600, 16gb ram, 500gb ssd, gtx 1080 ti
", "rawMarkdown": "ryzen 5 1600, 16gb ram, 500gb ssd, gtx 1080 ti", "votes": null }, { "id": "443761", "postDate": "12/22/2018 11:26:14", "content": "I use my computer i5, 14gb RAM, 2tb HDD, gtx 970 4gb for prototyping and testing some ideas.\nThen I run net on AWS server p2.xlarge with 12gb Tesla K80, 64gb RAM. But Im totally new to neural networks and this is my first competition, thats why strong hardware doesn't help me :D\nI have good ideas, but I just still have problems with understanding pytorch and how to run pytorch models through fastai
", "rawMarkdown": "I use my computer i5, 14gb RAM, 2tb HDD, gtx 970 4gb for prototyping and testing some ideas.\nThen I run net on AWS server p2.xlarge with 12gb Tesla K80, 64gb RAM. But Im totally new to neural networks and this is my first competition, thats why strong hardware doesn't help me :D\nI have good ideas, but I just still have problems with understanding pytorch and how to run pytorch models through fastai", "votes": null }, { "id": "443796", "postDate": "12/22/2018 13:01:32", "content": "i7-6850, 32GB RAM, 500GB SSD, 3x GTX 1080Ti
", "rawMarkdown": "i7-6850, 32GB RAM, 500GB SSD, 3x GTX 1080Ti", "votes": null }, { "id": "443830", "postDate": "12/22/2018 14:17:57", "content": "cool, 3x GTX 1080Ti\nI only have one GTX 1080Ti
", "rawMarkdown": "cool, 3x GTX 1080Ti\nI only have one GTX 1080Ti", "votes": null }, { "id": "443993", "postDate": "12/22/2018 21:36:42", "content": "i5-3570, 8GB RAM, 200GB SSD, GTX 1070
\n\nI'm in need of some upgrades :P
", "rawMarkdown": "i5-3570, 8GB RAM, 200GB SSD, GTX 1070\n\nI'm in need of some upgrades :P", "votes": null }, { "id": "443997", "postDate": "12/22/2018 22:01:49", "content": "Well, you are doing pretty well.... I detailed my build here: https://www.kaggle.com/getting-started/74592
", "rawMarkdown": "Well, you are doing pretty well.... I detailed my build here: https://www.kaggle.com/getting-started/74592", "votes": null }, { "id": "443999", "postDate": "12/22/2018 22:09:54", "content": "but I must say, I don't exclusively use the hardware for this competition. In parallel also for other competitions and work. I would say in average I dedicated 1.5 of the GPU for HPA but might be a bit more in the last two weeks.
", "rawMarkdown": "but I must say, I don't exclusively use the hardware for this competition. In parallel also for other competitions and work. I would say in average I dedicated 1.5 of the GPU for HPA but might be a bit more in the last two weeks.", "votes": null }, { "id": "444005", "postDate": "12/22/2018 22:27:59", "content": "Hopefully all the gamers will buy the new 2080ti for Christmas and the market will be flooded with 1080ti and I can get a second card.
", "rawMarkdown": "Hopefully all the gamers will buy the new 2080ti for Christmas and the market will be flooded with 1080ti and I can get a second card.", "votes": null }, { "id": "444069", "postDate": "12/23/2018 04:16:08", "content": "hope so, I also want to get another 1080Ti :)
", "rawMarkdown": "hope so, I also want to get another 1080Ti :)", "votes": null }, { "id": "444083", "postDate": "12/23/2018 05:42:04", "content": "When I started I was using:\n- i7-5930K\n- 64GB ram\n- 256GB sata ssd\n- GTX 970\n- Titan X Maxwell
\n\nToday it is at:\n- same cpu/ram\n- 512GB nvme ssd\n- RTX 2080\n- Titan X Maxwell
\n\nGetting Tensorflow to work with the RTX 2080 on Windows 7 took way too much work.
", "rawMarkdown": "When I started I was using:\n- i7-5930K\n- 64GB ram\n- 256GB sata ssd\n- GTX 970\n- Titan X Maxwell\n\nToday it is at:\n- same cpu/ram\n- 512GB nvme ssd\n- RTX 2080\n- Titan X Maxwell\n\nGetting Tensorflow to work with the RTX 2080 on Windows 7 took way too much work.", "votes": null }, { "id": "444128", "postDate": "12/23/2018 09:27:33", "content": "Mine: ryzen 5 1600, 32 gb ram, 500 gb ssd, gtx 1080i
", "rawMarkdown": "Mine: ryzen 5 1600, 32 gb ram, 500 gb ssd, gtx 1080i", "votes": null }, { "id": "444370", "postDate": "12/23/2018 22:10:43", "content": "Same old configuration since last two years:
\n\ni7 - 16 core\n32GB ram\n500 GB SSD\n1 x Titan X pascal
\n\nAs for deep learning hardware, I believe an investment in SSD is next to the GPU.
", "rawMarkdown": "Same old configuration since last two years:\n\ni7 - 16 core\n32GB ram\n500 GB SSD\n1 x Titan X pascal\n\nAs for deep learning hardware, I believe an investment in SSD is next to the GPU.", "votes": null }, { "id": "444377", "postDate": "12/23/2018 22:52:25", "content": "There were no major breakthroughs, so its still up to date. I might consider an m.2 ssd as my motherboard supports it. Will check if io is a bottleneck before investing...
", "rawMarkdown": "There were no major breakthroughs, so its still up to date. I might consider an m.2 ssd as my motherboard supports it. Will check if io is a bottleneck before investing...", "votes": null }, { "id": "444382", "postDate": "12/23/2018 23:09:18", "content": "I went from sata ssd to m.2 ssd. It helped somewhat but my machine is more CPU bound doing augmentation with 6 cores.
", "rawMarkdown": "I went from sata ssd to m.2 ssd. It helped somewhat but my machine is more CPU bound doing augmentation with 6 cores.", "votes": null }, { "id": "444619", "postDate": "12/24/2018 11:59:52", "content": "i5-8500(6 Core), 16GB RAM, GTX 1080, 2TB HDD(Yeah I'm about to add a SSD soon).
", "rawMarkdown": "i5-8500(6 Core), 16GB RAM, GTX 1080, 2TB HDD(Yeah I'm about to add a SSD soon).", "votes": null }, { "id": "445395", "postDate": "12/26/2018 11:26:45", "content": "i7 8 Core, 8 GB RAM, 1TB HDD. Macbook Pro
", "rawMarkdown": "i7 8 Core, 8 GB RAM, 1TB HDD. Macbook Pro", "votes": null }, { "id": "455046", "postDate": "01/12/2019 20:17:38", "content": "So, to conclude this very brief survey, now that we can see final results... hardware matters, it seems, to a point. You need enough hardware to perform training in a reasonable time, but its not enough - skill is as important. I would say Shai's setup is probably optimal. I would, perhaps, have another, cheaper card for performing inference and experiments in addition to the main one.
\n\nThank you for all participants!
", "rawMarkdown": "So, to conclude this very brief survey, now that we can see final results... hardware matters, it seems, to a point. You need enough hardware to perform training in a reasonable time, but its not enough - skill is as important. I would say Shai's setup is probably optimal. I would, perhaps, have another, cheaper card for performing inference and experiments in addition to the main one.\n\nThank you for all participants!", "votes": null } ], "comments": [ { "id": 443687, "author_name": "thegame76", "author_url": "", "post_date": "12/22/2018 06:32:31", "content": "ryzen 5 1600, 16gb ram, 500gb ssd, gtx 1080 ti
", "votes": null, "replies": [] }, { "id": 443761, "author_name": "baesiann", "author_url": "", "post_date": "12/22/2018 11:26:14", "content": "I use my computer i5, 14gb RAM, 2tb HDD, gtx 970 4gb for prototyping and testing some ideas.\nThen I run net on AWS server p2.xlarge with 12gb Tesla K80, 64gb RAM. But Im totally new to neural networks and this is my first competition, thats why strong hardware doesn't help me :D\nI have good ideas, but I just still have problems with understanding pytorch and how to run pytorch models through fastai
", "votes": null, "replies": [] }, { "id": 443796, "author_name": "christofhenkel", "author_url": "", "post_date": "12/22/2018 13:01:32", "content": "i7-6850, 32GB RAM, 500GB SSD, 3x GTX 1080Ti
", "votes": null, "replies": [ { "id": 443830, "author_name": "crilinux", "author_url": "", "post_date": "12/22/2018 14:17:57", "content": "cool, 3x GTX 1080Ti\nI only have one GTX 1080Ti
", "votes": null, "replies": [] }, { "id": 443999, "author_name": "christofhenkel", "author_url": "", "post_date": "12/22/2018 22:09:54", "content": "but I must say, I don't exclusively use the hardware for this competition. In parallel also for other competitions and work. I would say in average I dedicated 1.5 of the GPU for HPA but might be a bit more in the last two weeks.
", "votes": null, "replies": [] }, { "id": 444005, "author_name": "moshel", "author_url": "", "post_date": "12/22/2018 22:27:59", "content": "Hopefully all the gamers will buy the new 2080ti for Christmas and the market will be flooded with 1080ti and I can get a second card.
", "votes": null, "replies": [] }, { "id": 444069, "author_name": "crilinux", "author_url": "", "post_date": "12/23/2018 04:16:08", "content": "hope so, I also want to get another 1080Ti :)
", "votes": null, "replies": [] } ] }, { "id": 443993, "author_name": "pjbutcher", "author_url": "", "post_date": "12/22/2018 21:36:42", "content": "i5-3570, 8GB RAM, 200GB SSD, GTX 1070
\n\nI'm in need of some upgrades :P
", "votes": null, "replies": [ { "id": 443997, "author_name": "moshel", "author_url": "", "post_date": "12/22/2018 22:01:49", "content": "Well, you are doing pretty well.... I detailed my build here: https://www.kaggle.com/getting-started/74592
", "votes": null, "replies": [] } ] }, { "id": 444083, "author_name": "ldm314", "author_url": "", "post_date": "12/23/2018 05:42:04", "content": "When I started I was using:\n- i7-5930K\n- 64GB ram\n- 256GB sata ssd\n- GTX 970\n- Titan X Maxwell
\n\nToday it is at:\n- same cpu/ram\n- 512GB nvme ssd\n- RTX 2080\n- Titan X Maxwell
\n\nGetting Tensorflow to work with the RTX 2080 on Windows 7 took way too much work.
", "votes": null, "replies": [] }, { "id": 444128, "author_name": "zungmann", "author_url": "", "post_date": "12/23/2018 09:27:33", "content": "Mine: ryzen 5 1600, 32 gb ram, 500 gb ssd, gtx 1080i
", "votes": null, "replies": [] }, { "id": 444370, "author_name": "sgalib", "author_url": "", "post_date": "12/23/2018 22:10:43", "content": "Same old configuration since last two years:
\n\ni7 - 16 core\n32GB ram\n500 GB SSD\n1 x Titan X pascal
\n\nAs for deep learning hardware, I believe an investment in SSD is next to the GPU.
", "votes": null, "replies": [ { "id": 444377, "author_name": "moshel", "author_url": "", "post_date": "12/23/2018 22:52:25", "content": "There were no major breakthroughs, so its still up to date. I might consider an m.2 ssd as my motherboard supports it. Will check if io is a bottleneck before investing...
", "votes": null, "replies": [] }, { "id": 444382, "author_name": "ldm314", "author_url": "", "post_date": "12/23/2018 23:09:18", "content": "I went from sata ssd to m.2 ssd. It helped somewhat but my machine is more CPU bound doing augmentation with 6 cores.
", "votes": null, "replies": [] } ] }, { "id": 444619, "author_name": "mukul97", "author_url": "", "post_date": "12/24/2018 11:59:52", "content": "i5-8500(6 Core), 16GB RAM, GTX 1080, 2TB HDD(Yeah I'm about to add a SSD soon).
", "votes": null, "replies": [] }, { "id": 445395, "author_name": "pkurumbudel", "author_url": "", "post_date": "12/26/2018 11:26:45", "content": "i7 8 Core, 8 GB RAM, 1TB HDD. Macbook Pro
", "votes": null, "replies": [] }, { "id": 455046, "author_name": "moshel", "author_url": "", "post_date": "01/12/2019 20:17:38", "content": "So, to conclude this very brief survey, now that we can see final results... hardware matters, it seems, to a point. You need enough hardware to perform training in a reasonable time, but its not enough - skill is as important. I would say Shai's setup is probably optimal. I would, perhaps, have another, cheaper card for performing inference and experiments in addition to the main one.
\n\nThank you for all participants!
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "443092": "Just out of curiosity... Was wondering how strong is the correlation between computing power and LB score.\nMine is e5-1650, 16gb ram, 128gb ssd, 1tb sata, titan X (maxwell)", "443687": "ryzen 5 1600, 16gb ram, 500gb ssd, gtx 1080 ti", "443761": "I use my computer i5, 14gb RAM, 2tb HDD, gtx 970 4gb for prototyping and testing some ideas.\nThen I run net on AWS server p2.xlarge with 12gb Tesla K80, 64gb RAM. But Im totally new to neural networks and this is my first competition, thats why strong hardware doesn't help me :D\nI have good ideas, but I just still have problems with understanding pytorch and how to run pytorch models through fastai", "443796": "i7-6850, 32GB RAM, 500GB SSD, 3x GTX 1080Ti", "443830": "cool, 3x GTX 1080Ti\nI only have one GTX 1080Ti", "443993": "i5-3570, 8GB RAM, 200GB SSD, GTX 1070\n\nI'm in need of some upgrades :P", "443997": "Well, you are doing pretty well.... I detailed my build here: https://www.kaggle.com/getting-started/74592", "443999": "but I must say, I don't exclusively use the hardware for this competition. In parallel also for other competitions and work. I would say in average I dedicated 1.5 of the GPU for HPA but might be a bit more in the last two weeks.", "444005": "Hopefully all the gamers will buy the new 2080ti for Christmas and the market will be flooded with 1080ti and I can get a second card.", "444069": "hope so, I also want to get another 1080Ti :)", "444083": "When I started I was using:\n- i7-5930K\n- 64GB ram\n- 256GB sata ssd\n- GTX 970\n- Titan X Maxwell\n\nToday it is at:\n- same cpu/ram\n- 512GB nvme ssd\n- RTX 2080\n- Titan X Maxwell\n\nGetting Tensorflow to work with the RTX 2080 on Windows 7 took way too much work.", "444128": "Mine: ryzen 5 1600, 32 gb ram, 500 gb ssd, gtx 1080i", "444370": "Same old configuration since last two years:\n\ni7 - 16 core\n32GB ram\n500 GB SSD\n1 x Titan X pascal\n\nAs for deep learning hardware, I believe an investment in SSD is next to the GPU.", "444377": "There were no major breakthroughs, so its still up to date. I might consider an m.2 ssd as my motherboard supports it. Will check if io is a bottleneck before investing...", "444382": "I went from sata ssd to m.2 ssd. It helped somewhat but my machine is more CPU bound doing augmentation with 6 cores.", "444619": "i5-8500(6 Core), 16GB RAM, GTX 1080, 2TB HDD(Yeah I'm about to add a SSD soon).", "445395": "i7 8 Core, 8 GB RAM, 1TB HDD. Macbook Pro", "455046": "So, to conclude this very brief survey, now that we can see final results... hardware matters, it seems, to a point. You need enough hardware to perform training in a reasonable time, but its not enough - skill is as important. I would say Shai's setup is probably optimal. I would, perhaps, have another, cheaper card for performing inference and experiments in addition to the main one.\n\nThank you for all participants!" }, "source": "meta" }