Hey,
\n\nIt's taking days(weeks) for pre-trained networks to train on my laptop(Intel i3, 2.5 GHz, Quad core, 4 GB Ram). Can anyone suggest what specifications handle the data best?(Please mention the approximate training time as well )
\n\nThanks
", "messages": [ { "id": "119183", "postDate": "05/07/2016 20:51:14", "content": "Hey,
\n\nIt's taking days(weeks) for pre-trained networks to train on my laptop(Intel i3, 2.5 GHz, Quad core, 4 GB Ram). Can anyone suggest what specifications handle the data best?(Please mention the approximate training time as well )
\n\nThanks
", "rawMarkdown": "Hey,\r\n\r\nIt's taking days(weeks) for pre-trained networks to train on my laptop(Intel i3, 2.5 GHz, Quad core, 4 GB Ram). Can anyone suggest what specifications handle the data best?(Please mention the approximate training time as well ) \r\n\r\nThanks", "votes": null }, { "id": "119343", "postDate": "05/09/2016 09:43:05", "content": "Without a GPU you can forget the deep pre-trained models, they are too large and you would be looking at training times of months. It is realistic to try ZFTurbo's starter script and tune it a little. You should be able to get a score below 1.0 on your laptop with a few overnight training runs and some good choices when tuning the network. Chances of winning on any CPU-only rig are zero.
", "rawMarkdown": "Without a GPU you can forget the deep pre-trained models, they are too large and you would be looking at training times of *months*. It is realistic to try ZFTurbo's starter script and tune it a little. You should be able to get a score below 1.0 on your laptop with a few overnight training runs and some good choices when tuning the network. Chances of winning on any CPU-only rig are zero.", "votes": null }, { "id": "124893", "postDate": "06/23/2016 07:09:20", "content": "[quote=Neil Slater;119343]
\n\nWithout a GPU you can forget the deep pre-trained models, they are too large and you would be looking at training times of months. It is realistic to try ZFTurbo's starter script and tune it a little. You should be able to get a score below 1.0 on your laptop with a few overnight training runs and some good choices when tuning the network. Chances of winning on any CPU-only rig are zero.
\n\n[/quote]\nHi Neil . Can you tell me your computer rig for some competions that use cnn.
", "rawMarkdown": "[quote=Neil Slater;119343]\r\n\r\nWithout a GPU you can forget the deep pre-trained models, they are too large and you would be looking at training times of *months*. It is realistic to try ZFTurbo's starter script and tune it a little. You should be able to get a score below 1.0 on your laptop with a few overnight training runs and some good choices when tuning the network. Chances of winning on any CPU-only rig are zero.\r\n\r\n\r\n[/quote]\r\nHi Neil . Can you tell me your computer rig for some competions that use cnn.", "votes": null }, { "id": "124895", "postDate": "06/23/2016 07:11:34", "content": "[quote=panzer;124893]
\n\n[quote=Neil Slater;119343]
\n\nWithout a GPU you can forget the deep pre-trained models, they are too large and you would be looking at training times of months. It is realistic to try ZFTurbo's starter script and tune it a little. You should be able to get a score below 1.0 on your laptop with a few overnight training runs and some good choices when tuning the network. Chances of winning on any CPU-only rig are zero.
\n\n[/quote]\nHi Neil . Can you tell me your computer rig for some competions that use cnn.
\n\n[/quote]I want to buy a desktop and need a reference.Thx.
", "rawMarkdown": "[quote=panzer;124893]\r\n\r\n[quote=Neil Slater;119343]\r\n\r\nWithout a GPU you can forget the deep pre-trained models, they are too large and you would be looking at training times of *months*. It is realistic to try ZFTurbo's starter script and tune it a little. You should be able to get a score below 1.0 on your laptop with a few overnight training runs and some good choices when tuning the network. Chances of winning on any CPU-only rig are zero.\r\n\r\n\r\n[/quote]\r\nHi Neil . Can you tell me your computer rig for some competions that use cnn.\r\n\r\n[/quote]I want to buy a desktop and need a reference.Thx.", "votes": null }, { "id": "124951", "postDate": "06/23/2016 19:59:15", "content": "@panzer: I don't have a GPU and have dropped out of the competition. There are lots of choices, perhaps someone else who is competing on a GPU can give you some guidance.
", "rawMarkdown": "panzer: I don't have a GPU and have dropped out of the competition. There are lots of choices, perhaps someone else who is competing on a GPU can give you some guidance.", "votes": null }, { "id": "125008", "postDate": "06/24/2016 15:19:21", "content": "Hi panzer,
\n\nI use NVIDIA GTX 980 in this competition.
\nI recommend that you buy NVIDIA GTX 1080 which is newer products.
@Neil Slater: Thank you for the reply. It is a pity that you dropped out of the competion.
", "rawMarkdown": "Neil Slater: Thank you for the reply. It is a pity that you dropped out of the competion.", "votes": null }, { "id": "125075", "postDate": "06/25/2016 13:46:46", "content": "@toshi_k: Ok ,Thank you!
", "rawMarkdown": "toshi_k: Ok ,Thank you!", "votes": null } ], "comments": [ { "id": 119343, "author_name": "slobo777", "author_url": "", "post_date": "05/09/2016 09:43:05", "content": "Without a GPU you can forget the deep pre-trained models, they are too large and you would be looking at training times of months. It is realistic to try ZFTurbo's starter script and tune it a little. You should be able to get a score below 1.0 on your laptop with a few overnight training runs and some good choices when tuning the network. Chances of winning on any CPU-only rig are zero.
", "votes": null, "replies": [] }, { "id": 124893, "author_name": "lasshpanzer", "author_url": "", "post_date": "06/23/2016 07:09:20", "content": "[quote=Neil Slater;119343]
\n\nWithout a GPU you can forget the deep pre-trained models, they are too large and you would be looking at training times of months. It is realistic to try ZFTurbo's starter script and tune it a little. You should be able to get a score below 1.0 on your laptop with a few overnight training runs and some good choices when tuning the network. Chances of winning on any CPU-only rig are zero.
\n\n[/quote]\nHi Neil . Can you tell me your computer rig for some competions that use cnn.
", "votes": null, "replies": [] }, { "id": 124895, "author_name": "lasshpanzer", "author_url": "", "post_date": "06/23/2016 07:11:34", "content": "[quote=panzer;124893]
\n\n[quote=Neil Slater;119343]
\n\nWithout a GPU you can forget the deep pre-trained models, they are too large and you would be looking at training times of months. It is realistic to try ZFTurbo's starter script and tune it a little. You should be able to get a score below 1.0 on your laptop with a few overnight training runs and some good choices when tuning the network. Chances of winning on any CPU-only rig are zero.
\n\n[/quote]\nHi Neil . Can you tell me your computer rig for some competions that use cnn.
\n\n[/quote]I want to buy a desktop and need a reference.Thx.
", "votes": null, "replies": [] }, { "id": 124951, "author_name": "slobo777", "author_url": "", "post_date": "06/23/2016 19:59:15", "content": "@panzer: I don't have a GPU and have dropped out of the competition. There are lots of choices, perhaps someone else who is competing on a GPU can give you some guidance.
", "votes": null, "replies": [] }, { "id": 125008, "author_name": "toshik", "author_url": "", "post_date": "06/24/2016 15:19:21", "content": "Hi panzer,
\n\nI use NVIDIA GTX 980 in this competition.
\nI recommend that you buy NVIDIA GTX 1080 which is newer products.
@Neil Slater: Thank you for the reply. It is a pity that you dropped out of the competion.
", "votes": null, "replies": [] }, { "id": 125075, "author_name": "lasshpanzer", "author_url": "", "post_date": "06/25/2016 13:46:46", "content": "@toshi_k: Ok ,Thank you!
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "119183": "Hey,\r\n\r\nIt's taking days(weeks) for pre-trained networks to train on my laptop(Intel i3, 2.5 GHz, Quad core, 4 GB Ram). Can anyone suggest what specifications handle the data best?(Please mention the approximate training time as well ) \r\n\r\nThanks", "119343": "Without a GPU you can forget the deep pre-trained models, they are too large and you would be looking at training times of *months*. It is realistic to try ZFTurbo's starter script and tune it a little. You should be able to get a score below 1.0 on your laptop with a few overnight training runs and some good choices when tuning the network. Chances of winning on any CPU-only rig are zero.", "124893": "[quote=Neil Slater;119343]\r\n\r\nWithout a GPU you can forget the deep pre-trained models, they are too large and you would be looking at training times of *months*. It is realistic to try ZFTurbo's starter script and tune it a little. You should be able to get a score below 1.0 on your laptop with a few overnight training runs and some good choices when tuning the network. Chances of winning on any CPU-only rig are zero.\r\n\r\n\r\n[/quote]\r\nHi Neil . Can you tell me your computer rig for some competions that use cnn.", "124895": "[quote=panzer;124893]\r\n\r\n[quote=Neil Slater;119343]\r\n\r\nWithout a GPU you can forget the deep pre-trained models, they are too large and you would be looking at training times of *months*. It is realistic to try ZFTurbo's starter script and tune it a little. You should be able to get a score below 1.0 on your laptop with a few overnight training runs and some good choices when tuning the network. Chances of winning on any CPU-only rig are zero.\r\n\r\n\r\n[/quote]\r\nHi Neil . Can you tell me your computer rig for some competions that use cnn.\r\n\r\n[/quote]I want to buy a desktop and need a reference.Thx.", "124951": "panzer: I don't have a GPU and have dropped out of the competition. There are lots of choices, perhaps someone else who is competing on a GPU can give you some guidance.", "125008": "Hi panzer,\r\n\r\nI use [NVIDIA GTX 980][1] in this competition.